Methods and apparatus for recording reproducing, transmitting and/or receiving compressed data and recording medium therefor

0001This invention relates to methods and apparatus for recording and/or reproducing and/or transmitting and/or receiving data, for example bit-compressed digital audio signals or the like, and recording media therefor. More particularly, but not exclusively, the invention relates to an apparatus, technique and a recording medium for recording with a compression mode of plural bit rates.

0002Our United States Patents Nos. <patcit id="pcit0001" dnum="US5243588A"><text>US-A-5 243 588</text></patcit> and <patcit id="pcit0002" dnum="US5244705A"><text>US-A-5 244 705</text></patcit> disclose a technique consisting of bit-compressing input digital audio signals and recording the bit-compressed signals in a burst fashion with a pre-set data volume as a recording unit.

0003This technique resides in employing a magneto-optical disc as a recording medium and recording and/or reproducing AD (adaptive differential) PCM audio data as prescribed in audio data formats such as CD-1 (CD-Interactive) or CD-ROM XA. The AD-PCM audio data is recorded in a burst fashion on the magneto-optical disc with, for example, 32 sectors of the AD-PCM data and a few linking sectors for interleaving, forming one recording unit.

0004Several modes of the AD-PCM audio data may be selected in the recording and/or reproducing apparatus employing the magneto-optical disc. For example, a level A with a compression ratio twice that of the conventional compact disc (CD) and a sampling frequency of 37.8 kHz, a level B with a compression ration four times that of the conventional compact disc (CD) and a sampling frequency of 37.8 kHz and a level C with a compression ratio eight times that of the conventional compact disc (CD) and a sampling frequency of 18.9 kHz, are prescribed. That is, with the above-mentioned level B, the digital audio data is compressed to approximately one fourth of the original volume, with the replay time of the disc recorded with the level B mode being four times that with the standard CD format. This indicates that the apparatus may be reduced in size because a recording/reproducing time about as long as that achieved with a standard size disc of 12 cm in diameter may be achieved with a smaller sized disc.

0005However, since the rotational speed of the disc is the same as that of a standard CD, the amount of the compressed data obtained with the level B per a pre-set time is four times that of the standard CD. Consequently, the same compressed data is repeatedly read four times, in a time unit of, for example, a sector or a cluster, and only one of the four superimposed compressed data is transmitted to audio reproduction. Specifically, during scanning or tracking of a spiral recording track, a track jump to return to the starting track position is performed for each revolution in order to repeatedly track the same spiral track four times. This indicates that at least one of the four scans of compressed sound data from the repeated reading operations suffices. Hence, the disclosed technique is effective against errors caused by disturbances, and consequently may be especially suitable for use in connection with a small-sized portable type device.

0006Since, a semiconductor memory will be employed in the future as a recording medium, it would be desirable to achieve further bit compression in order to improve compression efficiency. Specifically, audio signals are recorded and/or reproduced using a so-called IC card (which includes one or more such semiconductor memories), on or from which bit-compressed data is recorded or reproduced, respectively.

0007With the IC card, increases in storage capacity or lower costs are achieved through progress in semiconductor technology. Notwithstanding such progress, when the IC card was first supplied to the market, the IC card was believed to have inadequate storage capacity, and furthermore, was expensive. Therefore, it may be contemplated to repeatedly transfer the recording contents of an inexpensive large-capacity recording medium, such as the magneto-optical disc, to the IC card by re-writing data a number of times. Specifically, desired musical numbers recorded on the magneto-optical disc are dubbed to the IC card and exchanged (or overwritten) with other musical numbers already stored in the IC card. By frequently re-writing the contents of the IC card in this manner, a wide variety of musical numbers may be enjoyed outdoors with a small number of IC cards which are on hand.

0008For recording/reproducing audio signals, however, the frequency bandwidths and the signal to noise characteristics differ for a variety of applications. For example, if high quality audio signals are desired, the bandwidth of from 15 kHz to 20 kHz and a good signal to noise ratio are required. For such purposes, a higher bit rate is tolerated. The bit rate in this case is usually 256 kbps to 64 kbps per channel.

0009Conversely, if, for the most part, audio signals are handled, a bandwidth of 5 to 7 kHz suffices, while a significantly high signal to noise ratio is not required. However, in order to increase recording/reproducing time for a given memory size, it is desirable to lower the bit rate from 64 kbps to several kbps. In light of the above, it is necessary to provide a recording/reproducing apparatus which is capable of accommodating the plural usages having different demand levels and in which it is possible to reduce the economic load to the maximum extent possible. However, if plural modes having different bandwidths have to be provided, it therefore is necessary to support plural sampling frequencies, with the result that the sampling frequency signal generating circuit tends to be complex or the LSI scale tends to be unavoidably increased. In addition, if the sampling frequency differs from one mode to another, it is difficult to transfer the information from one mode to another. If it is desired to write high bit rate mode information recorded on a large-capacity magneto-optical disc onto a small-capacity IC card at a low bit rate mode, it is necessary to completely annul the compression mode to restore the information to time-domain signals and to re-compress the signals at a low bit rate mode, with the consequence that the quantity of the arithmetic-logical steps is increased, thereby increasing the difficulty of real-time processing.

0010With the modes having increasingly low bit rates, however, the sound quality is lowered due to the decrease in the number of usable bits. If, when the bandwidth is narrowed, the bandwidth of the frequency division for compression is constant over the entire frequency range, the bandwidth resulting from division of the entire frequency range of 20 kHz into 32 bands of equal widths is 700 Hz, which is significantly wider than the low-range critical band of 100 Hz and wider than the critical bands for the major portion of the frequency range. This results in significantly lowered compression efficiency. However, if, when the bit rate is lowered, bit reduction is made preferentially for one of the main information and the subsidiary information, the sound quality is lowered significantly. For this reason, it becomes necessary to reduce not only the main information but also the subsidiary information as well.

0011A signal processing technique is described in <patcit id="pcit0003" dnum="GB2257606A"><text>GB-A-2,257,606</text></patcit> in which data recorded with bit compression at a constant bit rate on a magneto-optical disc are reproduced and transmitted to an excess bit removal circuit without an expanding operation, so that at least some bits which will give a quantization noise level imperceptible to human ears are removed by bit compression in relatively small blocks resulting from division along both the time axis and the frequency axis before recording the data on an IC card.

0012A method of recording compressed digital audio data by removing redundant bits is described in <patcit id="pcit0004" dnum="GB2258372A"><text>GB-A-2, 258,372</text></patcit>. In this technique, compressed audio digital data is recorded on a medium such as an IC card. The signal to be recorded is derived by an excess bit reduction circuit which removes redundant bits (which make no difference to quantization noise perceived by the human ear) from the compressed digital data, resulting in a recording signal of variable bit rate.

0013Various different aspects of the invention are set forth in the respective different independent claims of this specification.

0014Preferred embodiments of the invention described in detail below provide: <ul id="ul0001" list-style="none" compact="compact"><li>digital signal processing methods and apparatus in which a sampling frequency signal generating circuit is uncomplicated and the ISI scale need not be increased as a result of the provision of plural bit rate modes;</li><li>a compressed data recording/reproducing apparatus in which bit-compressed data from a recording media such as a magneto-optical disc or an optical disc may be dubbed with a reduced amount of arithmetic-logical operations or where the bit-compressed data may be reproduced from another recording medium such as an IC card with a reduced amount of arithmetic-logical operations; and</li><li>a compressed data recording/reproducing apparatus in which it is possible to suppress the volume of the subsidiary information to a necessary minimum value and to thereby decrease the deterioration in the sound quality due to a low bit rate.</li></ul>

0015With the preferred compressed data recording and/or reproducing apparatus and technique and recording medium, the same sampling frequency is employed, despite the difference in the bit rate for the respective modes, in order to avoid complexity in the structure of the sampling frequency generating circuit and to avoid an increase in hardware scale resulting from such complexity.

0016Moreover, the transfer of the information, which it has been difficult to achieve when using different sampling frequencies for the respective modes, may be easily achieved. For example, if it is desired to write the high bit rate mode information from a large-capacity magneto-optical disc onto a small-capacity IC card with a low bit rate mode, low bit rate compression may be realized by only an additional processing operation, without the necessity of annulling the compression mode for reversion to signals in the time domain, such that the volume of the processing operations may be minimized, while also achieving the real-time processing operation. The switching from the low bit rate mode to a bit rate mode with a bit rate at least higher than the low bit rate may be achieved only by format conversion, that is by rearraying the encoded information data.

0017According to a preferred feature of the invention, the conversion between different bit rate modes may be performed at a high speed without reversion to time-domain signals by employing the same orthogonal transform block size for respective bit rate modes. In addition, plural blocks for block floating and/or quantization noise affected blocks, which are adjacent to one another along the time scale or the frequency scale, may be used in common for diminishing the volume of the subsidiary information as required for the blocks for block floating and/or the quantization noise affected blocks, such as the scale factor or the word length data. Since acoustic signals exhibit high correlation along both the time scale and the frequency scale, there is little effect (in the case of acoustic signals) on the sound quality if the blocks for block floating and/or the quantization noise affected blocks are used in common. If the signals are temporally non-stationary, it is possible to employ a variable orthogonal transform block size for preventing the compression efficiency from being lowered. The bits for the subsidiary information thus deleted may be allocated for the main information. In addition, by changing the group of the subsidiary information to be used in common depending on the orthogonal transform block size, the orthogonal transform block size may be changed, such that the subsidiary information may be used in common in an optimum manner even although the structure of the blocks for block floating or the blocks affected by the quantization noise is changed.

0018On the other hand, if, in the case of audio signals, with the frequency bandwidth for quantization noise control being the same for all modes irrespective of the frequencies, the frequency rate of 20 kHz is divided into 32 bands, the frequency bandwidth is in the region of 700 Hz, which is significantly broader than the critical bandwidth of 100 Hz for the lower frequency range. Thus, the frequency bandwidth becomes narrower than the bandwidth of the critical band for the mid to low range, thus significantly lowering the quantization and encoding efficiency. Preferably, the frequency bandwidth for quantization noise control is selected to be broader towards the higher frequencies for at least the majority of the frequency bands resulting from division so that the frequency bandwidth for quantization noise control will be closer to the critical bandwidth.

0019When dubbing the bit-compressed data from the recording medium, such as the magneto-optical disc, onto another recording medium, such as an IC card, the data may be dubbed directly or after additional compression without at least performing full bit expansion. If the additional compression is to be performed, re-allocation and re-quantization of data bits may be performed in the frequency domain without signal transformation, with data bits being recorded along with the common plural subsidiary data.

0020With the compressed data recording and/or reproducing apparatus and technique and the recording medium, the same sampling frequency may be employed for preventing the sampling frequency generating circuit from being complicated as well as preventing the hardware scale from being increased. The transfer of the information between modes with different rates may be achieved easily without the necessity of complicated operations such as sampling rate conversion. If it is desired to write the high bit rate mode information recorded on a large-capacity magneto-optical disc onto a small-capacity IC card with a low bit rate mode, the low bit rate mode compression may be realized by an additional processing operation for minimizing the increase in the volume of the arithmetic operations, while the operation may also be achieved on the real-time basis. In addition, the lowering of the sound quality may be prevented for the low bit rate mode.

0021The invention will now be further described, by way of illustrative and nonlimiting example, with reference to the accompanying drawings, in which: <ul id="ul0002" list-style="none" compact="compact"><li><figref idref="f0001">Fig. 1</figref> is a block circuit diagram illustrating a recording and/or reproducing apparatus for compressed data embodying the invention;</li><li><figref idref="f0002">Fig. 2</figref> illustrates the recording contents of a magneto-optical disc and an IC card;</li><li><figref idref="f0002">Fig. 3</figref> is a schematic front view of an example of the apparatus shown in <figref idref="f0001">Fig. 1</figref>;</li><li><figref idref="f0003">Fig. 4</figref> is a block circuit diagram showing an embodiment of an encoding apparatus for implementing a high efficiency encoding technique for audio data used in the embodiment shown in <figref idref="f0001">Fig. 1</figref>;</li><li><figref idref="f0004">Fig. 5</figref> is a diagrammatic view of illustrating a block size for orthogonal transform for the embodiment shown in <figref idref="f0001">Fig. 1</figref>;</li><li><figref idref="f0005">Fig. 6</figref> is a block circuit diagram illustrating a structure for implementing bit allocation and an arithmetic-logical function;</li><li><figref idref="f0006">Fig. 7</figref> is a graph showing the spectrum of critical bands and bands which have been divided to take block floating into account;</li><li><figref idref="f0007">Fig. 8</figref> is a graph showing relative levels of a masking spectrum in relation to spectral components within a signal spectrum;</li><li><figref idref="f0008">Fig. 9</figref> is a graph showing a minimum audibility curve and the masking spectrum combined together;</li><li><figref idref="f0009">Fig. 10</figref> is a block circuit diagram showing a concrete arrangement for implementing a second bit allocation method;</li><li><figref idref="f0010">Figs. 11a and 11b</figref> are graphs showing the bit allocation and noise spectrum for a flatter signal spectrum with the second bit allocation method;</li><li><figref idref="f0011">Figs. 12a and 12b</figref> are graphs showing the bit allocation and noise spectrum for a signal spectrum exhibiting high signal spectrum tonality with the second bit allocation method;</li><li><figref idref="f0012">Fig. 13</figref> is a graph showing the relation between frequency and time for a processing block of 11.6 msec for an A mode, with the frequency range being divided into 52 bands including critical bands and bands for taking the block floating into account, with the block size for the orthogonal transform being of a long mode;</li><li><figref idref="f0013">Fig. 14</figref> is a graph showing the relation between frequency and time for a processing block of 11.8 msec for the A mode, with the frequency range being divided into 52 bands including critical bands and bands for taking the block floating into account, with the block size for the orthogonal transform being of a short mode;</li><li><figref idref="f0014">Fig. 15</figref> is a graph showing the relation between frequency and time for a processing block of 23.2 msec for the B mode, with the frequency range being divided into 52 bands including critical bands and bands for taking the block floating into account, with the block size for the orthogonal transform being of a long mode;</li><li><figref idref="f0015">Fig. 16</figref> is a graph showing the relation between frequency and time for a processing block of 23.2 msec for the B mode, with the frequency range being divided into 11 bands for a processing block of 23.2 msec for the B mode including critical bands and bands for taking the block floating into account, with the block size for the orthogonal transform being of a short mode;</li><li><figref idref="f0016">Fig. 17</figref> is a graph showing the relation between frequency and time for a processing block of 23.2 msec for the B mode including critical bands and bands for taking block floating into account, with the block size for the orthogonal transform being of both a short mode and a long mode;</li><li><figref idref="f0017">Fig. 18</figref> is a block diagram showing an embodiment of an adaptive bit allocation and encoding circuit for the B mode;</li><li><figref idref="f0018">Fig. 19</figref> is a block circuit diagram showing a specific arrangement for high speed conversion from the A mode to the B mode; and</li><li><figref idref="f0019">Fig. 20</figref> is a block circuit diagram showing a specific example of a decoding apparatus for implementing a high efficiency encoding technique for audio data embodying the invention.</li></ul>

0022Illustrative embodiments of the invention will now be described in detail with reference to the drawings.

0023<figref idref="f0001">Fig. 1</figref> shows a schematic arrangement (block circuit diagram) of a recording and/or reproducing apparatus for compressed data according to an embodiment of the invention.

0024The recording/reproducing apparatus shown in <figref idref="f0001">Fig. 1</figref> comprises a recording/reproducing unit for a magneto-optical disc 1 as one of the recording media and a recording unit for an IC card 2, as another recording medium, with the two units arranged into one system. When recording signals reproduced from the magneto-optical disc 1 by the reproducing portion of the recording/reproducing unit for the magneto-optical disc onto the IC card 2 (by the recording portion of the recording/reproducing unit for the IC card 2), reproduced compressed data read by an optical head 53 from the magneto-optical disc 1 of the reproducing side and sent to a decoder 71 for EFM (eight to fourteen modulation) demodulation. The decoder 71 provides deinterleaving and error correction, that is ATC audio data, which is sent to a memory 85 of the IC card recording unit. The ATC audio data is processed by adding a variable bit rate encoding window by an additional compressor 84. The additional compressor 84 performs an additional operation of entropy coding for the memory 85. The ATC audio data is then recorded on the IC card 2 via an IC card interfacing circuit 86. In this manner, the reproduced compressed data is transmitted to the recording system for recording on the IC card 2 in a compressed state prior to expansion by an ATC decoder 73.

0025Meanwhile, for usual audio reproduction, compressed data is read from the recording medium (magneto-optical disc 1) intermittently or in a burst fashion with a pre-set amount of data, for example, 32 + several sectors, as a unit, and is expanded for conversion into continuous audio signals. During dubbing, compressed data on the recording medium is continuously read and transmitted to the recording system for recording. In this manner, high-speed or short-term dubbing may be performed depending on the data compression ratio.

0026The arrangement shown in <figref idref="f0001">Fig. 1</figref> is now explained in further detail. In the magneto-optical disc recording/replay unit of the compressed data recording and/or reproducing apparatus, shown in <figref idref="f0001">Fig. 1</figref>, the magneto-optical disc 1, which is rotated by a spindle motor 51, is employed as a recording medium. To record data onto the magneto-optical disc 1, a modulated magnetic field corresponding to the recording data is applied by a magnetic head 54, while the magneto-optical disc 1 is irradiated with a laser light beam by the optical head 53. In reproducing the recorded data, a recording track on the magneto-optical disc 1 is traced by a laser light beam for reproduction photomagnetically.

0027The following description is directed mainly to the recording/replay apparatus. The optical head 53 is made up of a laser light source, such as a laser diode, optical components, such as a collimator lens, an objective lens, a polarized light beam splitter or a cylindrical lens, and a photodetector which includes a light receiving section having a pre-set pattern. The optical head 53 is placed at a position facing the magnetic head 54 with the magneto-optical disc 1 in-between. For recording data on the magneto-optical disc 1, the magnetic head 54 is driven by a head driving circuit 66 of a recording system, as explained later herein, for applying a modulated magnetic field corresponding to the recording data to the disc 1, while a laser light beam is radiated by the optical head 53 onto a target track of the magneto-optical disc 1 for thermo-magnetic recording in accordance with magnetic field modulation. In addition, the optical head 53 detects the reflected laser light from the target track in order to detect focusing error by an astigmatic method and in order to determine tracking error by a push-pull method. When reproducing data from the magneto-optical disc 1, the optical head 53, in addition to detecting the focusing error and/or the tracking error, also detects the difference in the angle of polarization (the Kerr rotation angle) of the reflected light from the target track in order to generate playback signals.

0028An output of the optical head 53 is supplied to an RF circuit 55 which extracts the above-mentioned focusing error signals and the tracking error signals from the output. These extracted signals are transmitted to a servo control circuit 56. The RF circuit 55 also converts the playback signals into bi-level signals which are supplied to a decoder 71 of the reproducing system as later explained herein.

0029The servo control circuit 56 is made up of, for example, a focusing servo control circuit, a tracking servo control circuit, a spindle servo control circuit, a thread servo control circuit and so forth. The focusing servo control circuit focusing-controls the optical system of the optical head 53 so that the focusing signal is reduced to zero. The tracking servo control circuit controls the optical system of the optical head 53 so that the tracking signal is reduced to zero. The spindle motor servo controlling circuit controls the spindle motor 51 for running the magneto-optical disc 1 into rotation at a pre-set rotational velocity, for example, at a constant linear velocity. The thread servo control circuit shifts the optical head 53 and the magnetic head 54 to a target track position of the magneto-optical disc 1 as designated by a system controller 57. The servo control circuit 56 transmits information indicating the operating states of the various components controlled by the servo control circuit 56 to the system controller 57.

0030A key input unit 58 and a display 59 are connected to the system controller 57 which controls the recording system and the reproducing system under an operating mode as designated by input information entered at the key input unit 58. The system controller 57 also supervises the recording position and the replay position on the recording track traced by the optical head 53 and the magnetic head 54 based upon the sector-based address information from Q-data, such as header time or sub-code data which is recorded on the magneto-optical disc 1. In addition, the system controller 57 causes the bit compression mode to be displayed on the display 59, based on the bit compression mode information at an encoder 63, as selected by the key input unit 58 (as later explained), and on the compression mode information in the replay data as obtained from the RF circuit 55 via a replay system (as later explained), while causing the replay time to be displayed on the display 59 based on the data compression ratio in the bit compression mode and the playback position information on the recording track.

0031For playback time display, the sector-based address information (absolute time information) reproduced from header time or sub-code Q data from the recording track of the magneto-optical disc 1 is multiplied by a reciprocal of the data compression ratio for the bit compression mode, such as 4 for the bit compression ratio of 1/4, to determine the actual time information, which is displayed on the display 9. Meanwhile, during recording, if the absolute time information is pre-recorded on the recording track of the magneto-optical disc 1, that is, if the magneto-optical disc 1 is pre-formatted, it is possible to display the actual recording time by reading the pre-formatted absolute time information and multiplying it with a reciprocal of the data compression ratio.

0032In a recording system of the present disc recording/reproducing apparatus, analog audio input signals AIN are supplied from an input terminal 60 via a low-pass filter 61 to an A/D converter 62 which quantizes the analog audio input signals AIN. The digital audio input signals from the A/D converter 62 are supplied to an adaptive transform coding (ATC) PCM encoder 63. Alternatively, digital audio input signals DIN from an input terminal 67 is supplied to the ATC encoder 63 via a digital input interfacing circuit 68. The ATC encoder 63 performs bit compression (data compression), in accordance with the various modes of the ATC system as shown in Table 1. Such bit compression is performed, on the digital audio PCM data (which is the above-mentioned input signal AIN as quantized by the A/D converter 62) at a pre-set transfer rate, which is the above-mentioned input signal AIN as quantized by the A/D converter 62. The ATC encoder 63 has its operating mode designated by the system controller 57. For the B-mode, as an example, the data is the compressed data having the sampling frequency of 44.1 kHz and the bit rate of 64 kbps (ATC data) and is supplied to a memory 64. The data transfer rate for the stereo B mode is reduced to one-eighth of the data transfer rate for the standard CD-DA format of 75 sectors per sec, or to 9.375 sectors per sec. <tables id="tabl0001" num="0001"><table frame="none"><title><b>TABLE 1</b></title><tgroup cols="4" rowsep="0"><colspec colnum="1" colname="col1" colwidth="14mm" /><colspec colnum="2" colname="col2" colwidth="25mm" /><colspec colnum="3" colname="col3" colwidth="24mm" /><colspec colnum="4" colname="col4" colwidth="68mm" /><thead valign="top"><row rowsep="1"><entry namest="col1" nameend="col4" colsep="0" rowsep="0" align="left"><b>SAMPLING FREQUENCY = 44.1 kHz</b></entry></row><row rowsep="1"><entry colsep="0" rowsep="0"><b>MODE</b></entry><entry colsep="0" rowsep="0" align="center"><b>BIT RATE</b></entry><entry colsep="0" rowsep="0" align="center"><b>BANDWIDTH</b></entry><entry colsep="0" rowsep="0" align="center"><b>MAXIMUM PROCESSING BLOCK LENGTH</b></entry></row><row rowsep="1"><entry colsep="0" rowsep="0" /><entry colsep="0" rowsep="0" align="center"><b>kbps/channel</b></entry><entry colsep="0" rowsep="0" align="center"><b>kHz</b></entry><entry colsep="0" rowsep="0" align="center"><b>msec</b></entry></row></thead><tbody><row><entry colsep="0">A</entry><entry colsep="0" align="center">128</entry><entry colsep="0">22</entry><entry colsep="0" align="char" char="." charoff="6">11.6</entry></row><row><entry colsep="0">B</entry><entry colsep="0" align="center">64</entry><entry colsep="0">13</entry><entry colsep="0" align="char" char="." charoff="6">23.2</entry></row><row><entry colsep="0">C</entry><entry colsep="0" align="center">32</entry><entry colsep="0">5.5</entry><entry colsep="0" align="char" char="." charoff="6">34.8</entry></row><row rowsep="1"><entry colsep="0" rowsep="0">D</entry><entry colsep="0" rowsep="0" align="center">16</entry><entry colsep="0" rowsep="0">3</entry><entry colsep="0" rowsep="0" align="char" char="." charoff="6">46.4</entry></row></tbody></tgroup></table></tables>

0033In the embodiment of <figref idref="f0001">Fig. 1</figref>, it is assumed that the sampling frequency of the A/D converter 62 is fixed at the sampling frequency of the standard CD-DA format (44.1 kHz), and that the sampling frequency is maintained in the ATC encoder 63 to perform bit compression. Since the lower the bit rate of the mode, the narrower becomes the signal pass band, the cut-off frequency of the low-pass filter 61 is adjusted in a corresponding manner. That is, it suffices to control the cut-off frequency of the low-pass filter 61 of the A/D converter 62 depending on the compression mode.

0034The memory 64 is a buffer memory having its data writing and data reading controlled by the system controller 57. The memory 64 transiently stores ATC data supplied from the ATC encoder 63 for recording on the disc when the necessity arises. That is, for the stereo B mode, the compressed audio data supplied from the ATC encoder 63 has its data transfer rate diminished to one-eighth of the data transfer rate for the standard CD-DA format of 75 sector per sec, or to 9.375 sectors per sec. The compressed data is written continuously on the memory 64. Although it suffices to record the compressed data (ATC data) at a rate of one per eight sectors, as mentioned above, such recording at every eight sectors is practically impossible, so that sector-continuous recording is performed, as later explained. This recording is performed in a burst fashion at the data transfer rate of 75 sectors per sec, which is the same as the standard CD-DA format, with a cluster made up of pre-set plural sectors, for example, 32 + several sectors, as a recording unit, with the interposition of a non-recording period. That is, in the memory 64, the stereo B mode ATC audio data, continuously written at a low transfer rate of 9.375 (= 75/8) sectors per sec associated with the bit compression mode, is read as the recorded data in a burst fashion at the above-mentioned transfer rate of 75 sectors per sec. The overall data transfer rate of the data read out and recorded, inclusive of the non-recording period, is the low rate of 9.375 sectors per sec. However, the instantaneous data transfer rate within the time period of the recording operation performed in a burst fashion is the above-mentioned standard data transfer rate of 75 sectors per sec. Therefore, if the disc rotating velocity is the same as that of the standard CD-DA format, that is a constant linear velocity, recording is made at the same recording density and with the same recording pattern as those of the CD-DA format.

0035The ATC audio data to be recorded is read from the memory 64 at the above-mentioned instantaneous transfer rate of 75 sectors per sec and supplied to an encoder 65. The recording unit by which the data string supplied from the memory 64 to the encoder 65 is a cluster consisting of plural sectors, such as 32 sectors, and several cluster-linking sectors arrayed ahead and at back of the cluster. The cluster-linking sectors are selected to be longer than an interleaving length at the encoder 65 so that data of neighboring clusters are not affected by data interleaving.

0036The encoder 65 performs encoding for error correction, such as parity appendage and data interleaving, or EFM encoding, on the recording data which is supplied from the memory 64 in the burst fashion. The recorded data encoded by the encoder 65 is then supplied to a magnetic head driving circuit 66. The magnetic head 54 is connected to and driven by the magnetic head driving circuit 66 to apply to the magneto-optical disc 1 a modulated magnetic field corresponding to the recorded data.

0037The system controller 57, in addition to controlling the memory 64 as described above, also controls the recording position in such a manner that the recorded data read in the burst fashion from the memory 64 is continuously recorded on the recording track of the magneto-optical disc 1. The recording position is controlled by supervising the recording position of the recorded data read in the burst fashion by the system controller 57 from the memory 64 and by supplying a control signal designating the recording position on the recording track of the magneto-optical disc 1 to the servo control circuit 56.

0038The reproducing system of the recording/reproducing unit for the magneto-optical disc 1 is hereinafter explained. The reproducing system is employed for reproducing the recorded data continuously recorded on the recording track of the magneto-optical disc 1. The reproducing system includes a decoder 71, to which is provided a playback output. The playback output is produced by tracing the recording track of the magneto-optical disc 1 with a laser light beam from the optical head 53. The reflected laser light beam is sensed by the optical head 53 and is converted into bi-level data by the RF circuit 55. Recorded data may be read not only from the magneto-optical disc 1 but also from the read-only (or playback-only) optical disc of the same type as the so-called compact disc.

0039The decoder 71 is a counterpart of the encoder 65 of the above-described recording system and performs the above-mentioned decoding for error correction or EFM decoding on the bi-level playback output of the RF circuit 55. The decoder 71 also reproduces the stereo B mode ATC audio data at a transfer rate of 75 sectors per second, which is faster than the normal transfer rate for the stereo B mode. The playback data obtained by the decoder 71 is supplied to a memory 72.

0040The memory 72 has its data writing and reading controlled by the system controller 57 in such a manner that the playback data supplied from the decoder 71 at the transfer rate of 75 sectors per sec is written therein in the burst fashion at the transfer rate of 75 sectors per sec. Such playback data, which is written in the burst fashion at the transfer rate of 75 sectors per sec in the memory 72, is continuously read from the memory 72 at the normal transfer rate for the stereo B mode of 9.375 sectors per sec.

0041The system controller 57 controls the memory 72 in such a manner as to write the replay data into the memory 72 at the transfer rate of 75 sectors per sec and to continuously read the replay data from the memory 72 at the above-mentioned transfer rate of 9.375 sectors per sec. The system controller 57, while controlling the memory 72 as described above, controls the replay position such that the replay data which is written in a burst fashion into the memory 72 under control by the system controller 57, is continuously reproduced from the recording track of the magneto-optical disc 1. The replay position is controlled such that the replay position for the replay data is supervised by the system controller 57 and a control signal designating the replay position on the recording track of the magneto-optical disc 1 or the optical disc is transmitted to the servo control circuit 56.

0042The stereo B mode ATC audio data obtained as the replay data continuously read from the memory 72 at the transfer rate of 9.375 sectors per sec is supplied to an ATC decoder 73. The ATC decoder 73 is a counterpart of the ATC encoder 63 of the recording system. The ATC decoder 73 and has its operating mode designated by the system controller 57 in such a manner that the stereo B mode ATC data, for example, is expanded by a factor of eight by way of bit expansion for reproducing 16-bit digital audio data. The digital audio data from the ATC decoder 73 is supplied to a D/A converter 74.

0043The D/A converter 74 translates the digital audio data supplied from the ATC decoder 73 into analog signals for forming analog audio output signals AOUT. The analog audio output signals AOUT are outputted at an output terminal 76 via a low-pass filter 75.

0044The above-mentioned IC card recording unit of the compressed data recording and/or reproducing apparatus is hereinafter explained. The analog audio input signals AIN from an input terminal 60 are supplied via a low-pass filter 61 to an A/D converter 62 for quantization. The digital audio signals from the A/D converter 62 are supplied to an additional compressor 84 which is a type of variable bit rate encoder performing a so-called entropy encoding. Thus, the entropy encoding by the additional compression unit 84 is performed as data is written in or read from the memory 85. The data from the additional compression unit 84, compressed and encoded at the variable bit rate for entropy coding, is recorded on the IC card 2 via an IC card interfacing circuit 86. It is, of course, possible with the present invention to effect a constant low bit rate recording by increasing the block size of the orthogonal transform or enlarging the frequency width of a frequency-domain block for block floating having the subsidiary information concerning the quantization and/or a block affected by the quantization noise instead of performing variable bit rate compression such as the entropy coding.

0045Meanwhile, the compressed data (ATC data) from the decoder 71 of the replay system of the recording/reproducing unit for the magneto-optical disc 1 is transmitted without expansion, that is, transmitted directly to the memory 85 of the IC card recording unit 4. Such data transfer is performed under control of the memory 85 by the system controller 57 during the so-called high-speed dubbing. It is also possible to transmit the compressed data from the memory 72 to the memory 85. Recording from the magneto-optical disc or the optical disc onto an IC card by changing the bit rate mode and by lowering the bit rate is suitable for recording on the IC card having high costs per a unit recording capacity. This is desireable because it is not accompanied by frequency conversion in which the sampling frequency need not be the same irrespective of the prevailing bit rate mode. Actual additional compression is performed in the additional compression unit 84.

0046The so-called high-speed digital dubbing operation is now explained. During high-speed digital dubbing, a dubbing key of the key input unit 58 is actuated for executing a pre-set high-speed dubbing under instructions issued from the system controller 57. Specifically, compressed data from decoder 71 is directly transmitted to the memory 85 of the IC card recording system and then processed with a variable bit rate encoding by the additional compression unit 84. The encoded data is then recorded on the IC card 2 via the IC card interfacing circuit 86. If the stereo B mode ATC audio data is recorded on the magneto-optical disc 1, data compressed by a factor of eight is continuously read from the decoder 71.

0047Consequently, during the high speed dubbing, compressed data corresponding to a duration eight times as long as the steady-state time duration on the real time basis for the stereo B mode is continuously produced from the magneto-optical disc 1. This compressed data is entropy coded or coded at a low constant bit rate for recording onto the IC card, so that a high dubbing speed eight times as high as the steady-state dubbing speed is achieved. Meanwhile, the multiplication factor in the high-speed dubbing differs with different compression modes. High-speed dubbing at a rate higher than the multiplication factor used for compression may also be performed, in which case the magneto-optical disc 1 is rotationally driven at a velocity a plurality of times faster than the steady-state velocity.

0048It is noted that not only is the data compressed and encoded at a constant bit rate recorded onto the magneto-optical disc 1, but also recorded is the information concerning the data volume at the time of variable rate compression and encoding of the data by the additional companding block 3, as shown in <figref idref="f0002">Fig. 2</figref>. By so doing, the number of the musical numbers recorded on the magneto-optical disc 1 and recordable on the IC card 2, as well as the combinations of the musical numbers recordable on the IC card 2, may be instantly known by reading the data volume information. The additional compression from the fixed bit rate to a lower bit rate mode may be performed in the additional companding block 84 in addition to entropy coding to the variable bit rate data.

0049Conversely, if the information concerning the data volume at the time of bit compression encoding at the constant bit rate, as well as the data compressed by bit compression and encoding at the variable bit rate, is recorded on the IC card 2, the data volume at the time of transmitting data, such as musical numbers, from the IC card 2 to the magneto-optical disc 1, may be known promptly. It is of course possible to record the data compressed and encoded at a fixed bit rate in the IC card 2 in addition to the data compressed and encoded at a variable bit rate.

0050<figref idref="f0002">Fig. 3</figref> shows, in a front view, the compressed data recording/reproducing apparatus 5, arranged and constructed as shown in <figref idref="f0001">Fig. 1</figref>. The apparatus 5 has an insertion section 6 for a magneto-optical disc or an optical disc and an insertion slot for an IC card 7. The disc and the IC card may be arranged separately from each other and interconnected by a cable (not shown).

0051The high efficiency compression and encoding is now explained in detail. That is, the technique of performing high efficiency encoding on input digital signals, such as audio PCM signals, by sub-band coding (SBC), adaptive transform coding (ATC) and adaptive bit allocation, is explained by referring first to <figref idref="f0003">Fig. 4</figref>.

0052In the high-efficiency encoding apparatus, shown in <figref idref="f0003">Fig. 4</figref>, the frequency range of the input digital signals is divided into plural frequency bands so that two neighboring lower most bands are of an equal bandwidth and the bandwidths become progressively broader in the direction of increasing frequencies. The input digital signals are orthogonally transformed from one frequency band to another to produce frequency-domain spectral data which is encoded with the numbers of bits which are adaptively allocated on the basis of so-called critical bands for the lower frequency range. Such adaptive allocation takes into account the characteristics of human aural sense, as later explained herein. In addition, smaller width bands are sub-divided from the critical bands for the higher frequency range in order to raise the block floating efficiency. This block usually becomes the block subject to the quantization noise, that is the block based on which the quantization noise is produced. In addition, with the present embodiment, the block size or block length is adaptively changed prior to orthogonal transform depending on input signals and floating is performed on the block basis.

0053Referring to <figref idref="f0003">Fig. 4</figref>, audio PCM signals having a frequency range of 0 to 22 kHz, with the sampling frequency of 44.1 kHz, are supplied to an input terminal 100. These input signals are divided by a frequency dividing filter 101, such as, for example, a QMF filter, into signals having a frequency range of 0 to 11 kHz and a frequency range of 11 to 22 kHz. The signals having the frequency range of 0 to 11 kHz are similarly divided by a similar frequency dividing filter 102, such as a QMF filter, into signals having a frequency band of 0 to 5.5 kHz and signals having a frequency band of 5.5 kHz to 11 kHz. The signals of the respective bands from the frequency dividing filters 101, 102 are supplied to an orthogonal transform block size decision circuit 106 for determining the block size on the band basis. With the orthogonal transform block size decision circuit 106, the block size length is based on, for example, the length of 11.6 ms which is the maximum block size. If the signals are temporally sub-stationary, the orthogonal transform block size is set to its maximum value of 11.6 ms for raising the frequency resolution. If the signals are temporally non-stationary, the orthogonal transform block size is further divided into four and eight sections for the frequency band not higher than 11 kHz and the frequency band higher than 11 kHz, respectively, for improving time resolution.

0054Among the techniques of dividing the above-mentioned input digital signals into plural frequency bands, there is a QMF filter, for example, as discussed in <nplcit id="ncit0001" npl-type="s"><text>Digital Coding of Speech in Sub-bands, R.E. Crochiere, Bell Syst. Tech. J., Vol.55 No.8, 1976</text></nplcit>. The technique of filter division into equal bandwidths is discussed in <nplcit id="ncit0002" npl-type="s"><text>Polyphase Quadrature Filters -A New Sub-band Coding Technique, Joseph H. Rothweiler, ICASSP 1983, Bost</text></nplcit>on.

0055Referring again to <figref idref="f0003">Fig. 4</figref>, outputs of the frequency dividing filters 101, 102 of the respective frequency bands are separately supplied to the orthogonal transform circuits 103, 104 and 105. The block sizes determined by the orthogonal transform size decision circuit 106 are supplied to the orthogonal transform circuits 103 to 105, outputs of which are blocked, that is formed into blocks, depending on the block sizes, for orthogonal transform. <figref idref="f0004">Fig. 5</figref> shows the orthogonal transform block sizes. Thus the block size of 11.6 ms (long mode) or the block size of 2.9 ms (short mode) is selected for the low to mid frequency range and the block size of 11.6 ms (long mode) or 1.45 ms (short mode) is selected for the high frequency range. The selected block size is outputted at a terminal 111 and thence supplied to the decoding circuit.

0056As examples of the above-described orthogonal transform, there is such orthogonal transform in which input audio signals are divided into time blocks of pre-set length (frames) and processed with fast Fourier transform (FFT), cosine transform (DCT) or modified DCT (MDCT) for transformation of time-domain signals into frequency-domain signals. Discussions on MDCT may be found in <nplcit id="ncit0003" npl-type="s"><text>Sub-band/ Transform Coding Using Filter Bank Designs Based on Time Domain Aliasing Cancellation, J. P. Princen, A. B. Bradley, ICASSP 1987, Univ. of Surrey, Royal Melbourne Inst. of Tech</text></nplcit>.

0057The bit allocation calculating circuit 107 calculates, in view of the masking effect, the amounts of masking for each critical band and each band divided from the critical band to take into account the block floating based on the spectral data divided to take into account the above-mentioned critical bands and block floating. In addition, the bit allocation calculating circuit 20 calculates the numbers of allocated bits for the respective bands, based upon the energies or peak values for each of the critical bands and the bands divided from the critical bands to take the block floating into account, and on the amounts of masking, and transmits the resulting information to the adaptive bit allocation and encoding circuit 108. The adaptive bit allocation and encoding circuit 108 requantizes the spectral data or the MDCT coefficient data depending on the numbers of bits allocated to the respective bands in the bit allocation calculating circuit 107. The data encoded in this manner is outputted at an output terminal 110.

0058<figref idref="f0005">Fig. 6</figref> shows, in a schematic block circuit diagram, an arrangement of a concrete embodiment of the bit allocation calculating circuit 107, in which the frequency-domain spectral data from the MDCT circuits 103 to 105 are supplied to an input terminal 21.

0059The frequency-domain input data is transmitted to a band-based energy calculating circuit 22 in which the energies of the critical bands and of the bands divided from the critical bands to take into account the block floating and the amounts of masking are found by calculating the sum total of the amplitudes in the respective bands. The amplitude peak values or mean values may also be employed in place of the energies in the respective bands. Each spectral component indicating the sum value of the amplitudes of each of the respective bands is indicated as SB in <figref idref="f0006">Fig. 7</figref> as an output of the energy calculating circuit 22. In <figref idref="f0006">Fig. 7</figref>, <figref idref="f0011">12</figref> bands B1 to B12 are shown as indicating the critical bands and the bands sub-divided from the critical bands to take into account the block floating, and the amounts of masking, for simplifying the drawing.

0060It is noted that an operation of multiplying each spectral component SB by a pre-set weighting function for taking into account the effects of masking is now performed by way of convolution. To this end, an output of the band-based energy calculating circuit 22, that is, each value of the spectral component SB, is transmitted to a convolution filter circuit 23. The convolution filter circuit 23 is made up of a plurality of delay elements for sequentially delaying input data, a plurality of multipliers, such as 25 multipliers associated with the respective bands, for multiplying outputs of the delay elements with filter coefficients or weighting functions, and an adder for finding the sum of the outputs of the respective multipliers. By such convolution, the sum of the portions indicated by broken lines in <figref idref="f0006">Fig. 7</figref> is found. "Masking" means the phenomenon in which certain signals are masked by other signals and become inaudible due to characteristics of the human aural sense. The masking effect may be classified into time-domain masking effect produced by the time-domain audio signals and concurrent masking effect produced by the frequency-domain signals. By this masking, any noise present in a masked portion becomes inaudible. In actual audio signals, the noise within the masked range is an allowable noise.

0061By way of a concrete example of multiplication coefficients or filter coefficients of the respective filters of the convolution filter circuit 23, if the coefficient of a multiplier M for an arbitrary band is 1, outputs of the delay elements are multiplied by coefficients 0.15, 0.0019, 0.0000086, 0.4, 0.06 and 0.007 at the multipliers M-1, M-2, M-3, M+1, M+2 and M+3, M being an arbitrary integer of from 1 to 25, for performing convolution of the spectral components SB.

0062An output of the convolution filter circuit 23 is transmitted to a subtractor 24 for finding a level α corresponding to the allowable noise level in the convolved region. Meanwhile, the allowable noise level α is such a level which will give an allowable noise level for each of the critical bands by deconvolution as will be described subsequently. The subtractor 24 is supplied with an allowance function (a function representative of the masking level) for finding the level α. The level α is controlled by increasing or decreasing the allowance function. The allowance function is supplied from a (n - ai) function generator 25 which is now explained.

0063That is, the level α corresponding to the allowable noise level is found from the equation (1):<maths id="math0001" num="(1)"><math display="block"><mi mathvariant="bold">α</mi><mo mathvariant="bold">=</mo><mi mathvariant="bold">S</mi><mo mathvariant="bold">-</mo><mfenced><mi mathvariant="bold">n</mi><mo mathvariant="bold">-</mo><mi mathvariant="bold">ai</mi></mfenced></math><img file="EP1517324B1_D0001.tif" /></maths> where <u>i</u> is the number accorded sequentially to the critical bands beginning from the lower side, <u>n</u> and <u>a</u> are constants where a > 0 and S the intensity of the convolved Bark spectrum. In the equation (1), (n - ai) represents the allowance function. In the present embodiment, by setting so that n = 38 and a = 1, optimum encoding may be achieved without deterioration in the sound quality.

0064The level α is determined in this manner and transmitted to a divider 26 for deconvolving the level α in the convolved region. By this deconvolution, the masking spectrum is found from the level α. This masking spectrum becomes the allowable noise level. Although the deconvolution necessitates complex arithmetic-logical steps, it is performed in the present embodiment in a simplified manner by using the divider 26.

0065The masking spectrum is transmitted via a synthesizing circuit 27 to a subtractor 28 which is supplied with an output of the band-based energy detection circuit 22, that is the above-mentioned spectral components SB. The subtractor 28 subtracts the masking spectrum from the spectrum SB for masking the portions of the spectral components SB lower than the level of the masking spectrum MS, as shown in <figref idref="f0007">Fig. 8</figref>.

0066An output of the subtractor 28 is outputted via an allowable noise correction circuit 30 at an output terminal 31 and thence transmitted to a ROM, not shown, in which the information concerning the number of the allocated bits is stored previously. The ROM outputs the information concerning the number of allocated bits for each band, depending on an output of the subtraction circuit 28 supplied via an allowable noise correction circuit 30, that is, depending on the level of a difference between the energies of the respective bands and outputs of the noise level setting means. The information concerning the number of the allocated bits is transmitted to the adaptive bit allocation and encoding circuit 108 to permit the frequency-domain spectral data from the MDCT circuits 103 to 105 to be quantized with the numbers of bits allocated to the respective bands.

0067In sum, the adaptive bit allocation and encoding circuit 108 quantizes the band-based spectral data with the numbers of bits allocated depending on the amounts of masking and the level of the difference between the output of the noise level setting means and the energies in the critical bands and the bands divided from the critical bands to take the block floating into account. The delay circuit 29 is provided for delaying the spectral components SB from the energy detection circuit 22 in consideration of the delay produced in the circuitry upstream of the synthesizing circuit 27.

0068The synthesizing circuit 27 synthesizes the masking spectrum MS and data from the minimum audibility curve RC from the minimum audibility curve generating circuit 32 representing characteristics of the human aural sense as shown in <figref idref="f0008">Fig. 9</figref>. If the absolute noise level is lower than the minimum audibility curve, the noise becomes inaudible. The minimum audibility curve differs with the difference in the replay level even although the coding is made in the same manner. However, since there is no marked difference in the manner of entering into the 16-bit dynamic range in music in actual digital systems, it may be presumed that, if the quantization noise of the frequency range in the vicinity of 4 kHz, which frequency is most perceptible to the ear is not heard, the quantization noise lower than the level of the minimum audibility curve is not heard in any other frequency range. Assuming that the recording/replay apparatus is employed so that the noise in the vicinity of 4 kHz ( for a word length employed by the system) is not heard, and the allowable noise level is to be obtained by synthesizing the minimum audibility curve RC and the masking spectrum MS, the allowable noise level may be up to the level indicated by hatched lines in <figref idref="f0008">Fig. 9</figref>. In the present embodiment, the level of 4 kHz of the minimum audibility curve is matched to the minimum level corresponding to, for example, 20 bits. In <figref idref="f0008">Fig. 9</figref>, the signal spectrum SS is also shown.

0069In addition, the allowable noise correction circuit 30 corrects the allowable noise level in the output of the subtractor 28 based on the information of the equal-loudness curve transmitted from a correction information outputting circuit 33. The equal-loudness curve is a characteristic curve concerning characteristics of human aural sense, and is obtained by finding the sound pressures of the sound at the respective frequencies heard with the same loudness as the pure tone of 1 kHz and by connecting the sound pressures by a curve. It is also known as an equal loudness sensitivity curve. The equal-loudness curve also delineates a curve which is substantially the same as the minimum audibility curve shown in <figref idref="f0008">Fig. 9</figref>. With the equal-loudness curve, the sound in the vicinity of 4 kHz is heard with the same loudness as the sound of 1 kHz, even although the sound pressure is decreased by 8 to 10 db from the sound of 1 kHz. Conversely, the sound in the vicinity of 50 kHz cannot be heard with the same loudness as the sound of 1 kHz unless the sound pressure is higher by about 15 db than that of the sound of 1 kHz. Thus it may be seen that the noise in excess of the level of the minimum audibility curve and within the allowable noise level preferably has frequency characteristics represented by a curve conforming to the equal-loudness curve. Thus it may be seen that correction of the allowable noise level in view of the equal-loudness curve is in conformity to the characteristics of the human aural sense.

0070Meanwhile, the correction information outputting circuit 33 may be designed to correct the allowable noise level based on the information of the error between the detection output of the volume of the output information (data quantity) at the time of quantization at the adaptive bit allocation and encoding circuit 108 and the target bit rate of the ultimately encoded data. There are occasions where the total number of bits as obtained by temporary adaptive bit allocation for the totality of the bit allocation unit blocks is different from the pre-set number of bits (target number of the bits) as determined by the bit rate of the ultimate encoded data and hence the bit allocation is again made so that the difference becomes equal to zero. That is, if the total number of allocated bits is less than the target number of the bits, the bits corresponding to the difference are additively allocated to respective unit blocks, whereas, if the total number of allocated bits is larger than the target number of the bits, the bits corresponding to the difference are deductively allocated to the respective unit blocks.

0071To this end, an error in the total number of allocated bits from the target number is detected and correction data for correcting the number of allocated bits is outputted by the correction information output circuit 33 depending on the error data. If the error data indicates the shortage of the number of bits, it is probable that more bits are employed per unit block so that the amount of the data is in excess of the target number of bits. If the error data indicates an excess of the number of bits, it is probable that a smaller number of bits per unit block suffices so that the amount of the data is less than the target number. Consequently, the correction information outputting circuit 33 is responsive to the error data to output the data of correction values by which the allowable noise level at an output of the subtractor 28 is corrected based on the information of the equal-loudness curve. The correction values are transmitted to the allowable noise correction circuit 30 for correcting the allowable noise from the subtractor 28. With the above-described system, data obtained by processing the orthogonal transform output spectrum by the subsidiary information for quantization is obtained as the main information, while the scale factor indicating the state of block floating or the word length data is produced as subsidiary information for quantization. The main information and the subsidiary information are transmitted from the encoder to the decoder.

0072The bit allocation calculating circuit 107 is arranged as shown for example in <figref idref="f0009">Fig. 10</figref>. The following effective technique of bit allocation, which is different from the above-described technique of bit allocation, is explained by referring to <figref idref="f0009">Fig. 10</figref>.

0073Outputs of the MDCT circuits 103 to 105 in <figref idref="f0003">Fig. 4</figref> are transmitted via an input terminal 300 of <figref idref="f0009">Fig. 10</figref> to an energy calculating circuit 301 adapted for calculating the band-based energies. In the band-based energy calculating circuit 301, the signal energies for each of the critical bands and each of the sub-divided bands for higher frequencies are found by calculating root mean square values of the amplitudes in the respective bands. Meanwhile, the peak values or mean values of the amplitudes may be employed in place of the band-based energies.

0074The band-based spectral components of the sums of the respective critical bands or the bands sub-divided from the critical bands for the higher frequencies, as outputted from the energy calculating circuit 301, become the spectral components SB (Bark spectrum) shown in <figref idref="f0006">Fig. 7</figref>.

0075If, in the present embodiment, the number of bits indicating MDCT coefficients employed for transmission or recording is 1 kbits/sec, a fixed bit allocation pattern is formed in the present embodiment using the 1 kbits. In the present embodiment, plural bit allocation patterns for the fixed bit allocation are provided and may be selectively employed depending on the signal properties. In the present embodiment, the fixed bit allocation circuit 307 has a variety of patterns in which the bits of short time blocks corresponding to 1 kbits are distributed for the respective frequencies. The fixed bit allocation circuit 307 has plural patterns having different bit allocation ratios for the mid to low frequencies and for the high frequencies. Such a pattern is selected which has the smaller number of bits allocated to the high frequencies for smaller signal magnitudes. In this manner, it becomes possible to take advantage of the loudness effect in which the sensitivity for the higher frequencies becomes lower for smaller signal magnitudes. Although the magnitude of the full-range signals may be employed as the signal magnitude, it is also possible to utilize an MDCT output or an output of the non-blocking frequency dividing circuit comprising, for example, a filter. The number of usable bits of 1 kbits, is set in, for example, a usable total bit number inputting circuit 305. The total number of bits may be entered from outside.

0076The output of the energy calculating circuit 301 is also transmitted to an energy dependent bit allocation circuit 306 which finds the energy-dependent bit allocation pattern from the band-based energies. The energy-dependent bit patterns are so set that the larger the energies of a given band, the greater the number of bits allocated to the band.

0077Referring to <figref idref="f0009">Fig. 10</figref>, the distribution ratio between the fixed bit allocation pattern and the bit allocation dependent upon the Bark spectrum (spectrum SB) is set by an index indicating the smoothness of the signal spectrum. That is, in the present embodiment, an output of the energy calculating circuit 301 is provided to a spectrum smoothness calculating circuit 302 in which the sum of absolute values of the differences between adjacent values of the signal spectral components divided by the sum of the signal spectral components is found and employed as the index of signal spectrum smoothness (tonality). Once this tonality is determined, the above-mentioned distribution ratio is found in the bit distribution ratio decision circuit 304. Meanwhile, the distribution ratio is used for changing the weighting ratio between the fixed bit allocation and the energy-dependent energy distribution.

0078The bit distribution ratio data from the bit distribution ratio decision circuit 304 is provided to a multiplier 312. The multiplier 312 is also provided with an output of the fixed bit distribution circuit 307, while being also supplied to a multiplier 311 supplied with an output of the energy dependent bit distribution circuit 306. Outputs of these multipliers 311, 312 are supplied to a summing circuit 308. That is, the number of bits determined by the fixed bit distribution and the number of bits by the spectrum-dependent bit distribution (based on energies of the critical bands and the bands sub-divided from the critical bands for the higher frequencies) are each multiplied by the values of the above-mentioned distribution ratio and the resulting products are summed by the summing circuit 308. The resulting sum data being supplied via an output terminal 309 (output terminal of band-based bit allocation) to downstream circuitry to be used for quantization.

0079The manner of bit allocation at this point is shown in <figref idref="f0010">Figs. 11b</figref> and <figref idref="f0011">12b</figref>. The state of the associated quantization noise is shown in <figref idref="f0010">Figs. 11a</figref> and <figref idref="f0011">12a</figref>. <figref idref="f0010">Figs. 11a and 11b</figref> illustrate a flatter signal spectrum, whereas <figref idref="f0011">Figs. 12a and 12b</figref> illustrate the signal spectrum exhibiting high tonality. In <figref idref="f0010">Figs. 11b</figref> and <figref idref="f0011">12b</figref>, Q_S indicates the signal level dependent bit quantity, and Q_F indicates the bit quantity corresponding to fixed bit allocation. In <figref idref="f0010">Figs. 11a</figref> and <figref idref="f0011">12a</figref>, L indicates the signal level and N_S and N_F indicate the signal level dependent noise reduction and the noise level due to fixed bit allocation, respectively.

0080In <figref idref="f0010">Fig. 11b</figref> showing the flatter signal spectrum, bit allocation due to a larger quantity of the fixed bit allocation is useful in assuring a larger signal to noise ratio over the entire frequency range. However, in <figref idref="f0010">Fig. 11b</figref>, a smaller number of bits are allocated for the low and high frequency ranges because these ranges are less critical to the aural sense. By signal level dependent allocation of a smaller number of bits for the low and high frequencies, as shown at Q_S in <figref idref="f0010">Fig. 11</figref>, the noise level of the bands having larger signal magnitudes is selectively reduced. Consequently, the frequency range selectivity becomes broader for a flatter signal spectrum.

0081Conversely, if the signal spectrum indicates high tonality, as shown in <figref idref="f0011">Fig. 12b</figref>, a large number of bits by the signal level dependent bit allocation is employed to reduce the quantization noise of an extremely narrow band indicated by N_S in <figref idref="f0011">Fig. 12a</figref>. This improves characteristics of lone spectral input signals. Also, the noise level may be non-selectively reduced over a wide frequency range by the smaller number of bits by the fixed bit allocation.

0082Returning to <figref idref="f0003">Fig. 4</figref>, the adaptive bit allocation and encoding circuit 108 is explained. In the present embodiment, there are two bit rate modes A and B, for example, with the bit rate mode A being 128 kbps/ channel and the bit rate mode B being one-half the bit rate for the mode A, or 64 kbps/ channel. However, there may also be more than two bit rate modes.

0083First, the encoding method for the mode A (A-mode) is explained. <figref idref="f0012">Figs. 13</figref> and <figref idref="f0013">14</figref> show a typical example of the block floating band division for the A-mode. In <figref idref="f0012">Fig. 13</figref>, in which the orthogonal transform block size is 11.6 ms, and in <figref idref="f0013">Fig. 14</figref>, in which the frequency range is divided into four and eight for the low to mid frequency range and for the high frequency range, respectively, the total number of the block floating bands is the same and equal to 52. As for the numbers of the block floating bands, as outputs of the frequency dividing filters 101 and 102, there are 20 block floating bands for the low frequency range and 16 block floating bands for the mid to high frequency range. Since these numbers remain fixed without regard to the orthogonal transform block size, there is no convenience even although the orthogonal transform block size is independently changed from band to band. For example, assuming that the block size is 1.6 ms divided by four only for the low frequency range, with the block size being 11.6 ms for the mid to high frequency range, the total number of the block floating bands becomes equal to 52 if the low frequency range and the mid to high frequency range are divided as shown in <figref idref="f0013">Figs. 14</figref> and <figref idref="f0012">13</figref>, respectively. In the adaptive bit allocation encoding circuit 108, there are accorded the information data for the scale factor and the word length for each of the 52 block floating bands, with the spectral data being quantized and encoded depending on the accorded scale factors and word lengths.

0084The encoded data is outputted at the output terminal 110 for recording or transmission.

0085Next, the encoding method for the B-mode is explained. Since the bit rate for the B-mode becomes one-half that of the A-mode, the volume of the subsidiary information, such as the scale factor, word length or the like, is not changed, and only the volume of the main information, that is spectral data, is diminished, if the encoding method remains the same as that used for the B-mode. That is, with the B-mode, the proportion of the subsidiary information in the total information volume is increased and hence the encoding efficiency is lowered as compared to the A-mode. If the bit rate is halved, it is desirable that not only the volume of the main information but also the volume of the subsidiary information be diminished to one half or less. In the present embodiment, reduction of the volume of the subsidiary information is achieved by having the value of the subsidiary information between two temporally adjacent blocks. That is, for halving the volume of the subsidiary information for the B-mode as compared to that for the A-mode. Since the volume of the subsidiary information for the A-mode is basically the same as that of the number of the block frequency bands, it is equal to 52/11.6 ms. For the B-mode, since the block floating band is extended along the time scale is extended, the volume of the subsidiary information is equal to 52 /23.2 ms. Thus the volume of the subsidiary information within the same time interval for the B-mode is half that for the A-mode. <figref idref="f0014 f0015 f0016">Figs. 15 to 17</figref> illustrate a an example of block floating band division for the B-mode.

0086In the graph of <figref idref="f0014">Fig. 15</figref>, the orthogonal transform block sizes of two temporally neighboring blocks are of the long mode. The areas delimited by solid lines each indicate an orthogonal transform block, while hatched areas each indicate a single block floating band or an area represented by one and the same subsidiary information.

0087That is, the block floating bands shown in <figref idref="f0014">Fig. 15</figref> are formed by unifying two block floating bands neighboring along the time scale for the A-mode as shown in <figref idref="f0012">Fig. 13</figref>. Stated differently, the subsidiary information as set for two temporally neighboring block floating bands for the A-mode in <figref idref="f0012">Fig. 13</figref> is used in common so that the same subsidiary information is set for these two block floating bands.

0088If viewed along the frequency scale, the relation between the block floating bands and the subsidiary information is the same as that shown in <figref idref="f0012">Fig. 13</figref>.

0089In the graph of <figref idref="f0015">Fig. 16</figref>, both the orthogonal transform block sizes of two temporally neighboring blocks are of the short modes. The areas delimited by solid lines each indicate an orthogonal transform block, while hatched areas each indicate a single block floating band. That is, the block floating bands shown in <figref idref="f0015">Fig. 16</figref> are formed by unifying two block floating bands neighboring along the time scale for the A-mode as shown in <figref idref="f0013">Fig. 14</figref>. The manner of frequency division into bands along the frequency scale is the same as that shown in <figref idref="f0013">Fig. 14</figref>.

0090In the graph of <figref idref="f0016">Fig. 17</figref>, the orthogonal transform block size differs between two temporally neighboring blocks, that is, the orthogonal transform block sizes of these blocks consist in the combination of the short and long modes. The areas delimited by solid lines each indicate an orthogonal transform block, while hatched areas each indicate a single block floating band. The blocks having the orthogonal transform block size of the short mode, that is the blocks for the mid range of 0 to 11.6 ms and the low and high ranges of 11.6 to 23.2 ms in <figref idref="f0016">Fig. 17</figref>, are the same as those of the short mode shown in <figref idref="f0015">Fig. 16</figref>.

0091In further detail, the two bands temporally neighboring along the time scale of the block floating band of the A-mode shown in <figref idref="f0013">Fig. 14</figref> are grouped together. In other words, the subsidiary information as set in the two temporally neighboring block floating bands is used in common, so that the same subsidiary information is set for the two block floating bands.

0092If viewed along the frequency scale, the manner of frequency division into bands and the relation between the block floating bands and the subsidiary information remain the same as those shown in <figref idref="f0013">Fig. 14</figref>.

0093As for the blocks having the orthogonal transform block size of the long mode, that is the blocks for the low and high ranges of 0 to 11.6 ms and the mid range of 11.6 to 23.2 ms as shown in <figref idref="f0016">Fig. 17</figref>, since the block floating bands in these blocks are not divided along the time scale, it being impossible to unify the bands along the time scale or to use the common subsidiary data between two neighboring block floating bands, the two bands neighboring each other along the frequency scale are unified into one band as exceptional measures. Alternatively, the subsidiary information as set for two block floating bands neighboring each other along the frequency scale is used in common so that the same subsidiary information is set for these two block floating bands. If viewed along the time scale, the relation between the block floating bands and the subsidiary information is the same as that shown in <figref idref="f0012">Fig. 13</figref>.

0094Thus, for halving the number of the subsidiary data for the B-mode as compared to that for the A-mode, the two block floating bands neighboring each other along the time scale or the frequency scale are used in common, or the subsidiary information as set for the two block floating bands neighboring each other along the time scale or the frequency scale is used in common. This allows the same subsidiary information to be used for these two block floating bands. As a result, the main information may be prevented from being decreased significantly with reduction in the bit rate, while the encoding efficiency may be improved.

0095<figref idref="f0017">Fig. 18</figref> shows an example of an adaptive bit allocation and encoding circuit 108 for the B-mode, in which the information of the orthogonal transform block size and spectral data for the above-mentioned spectral data (MDCT coefficients) are supplied to terminals 401 and 402, respectively. The information concerning the orthogonal transform block size is sent to a scale factor re-setting circuit 405 and a word length re-setting circuit 406, while the spectral data (MDCT coefficients) are sent to a quantizer 408.

0096The scale factors A for the temporally neighboring two block floating bands, as set for the respective bands at the time of block floating band division for the A-mode and supplied via an input circuit 403, are grouped together by the scale factor re-setting circuit 405, so that the scale factor B for the B-mode is re-set. Usually, the larger one of the two scale factors A is selected and used as a common scale factor.

0097Similarly, the word length B for the B-mode is re-set by the word length re-setting circuit 406 from the word lengths A which is set for the respective bands at the time of division into block floating bands and which is supplied via an input circuit 404. In re-setting the word length, the larger one of the two word lengths A, for example, is selected. Alteratively, a mean value of the two word lengths A may be employed.

0098Meanwhile, two blocks (23.2 ms) each of the scale factor A and the word length A are each sent as one unit to the re-setting circuits 405 and 406, respectively.

0099The re-set word length data is then corrected by the word length re-setting circuit 406 for errors produced by the re-setting operation in connection with the total numbers of bits. The scale factor B and the word length B as re-set are sent to a quantizer 408 and an encoder 409 so as to be employed during quantization of the spectral data. The quantized and encoded spectral data are outputted as encoded data B at a terminal 410.

0100In the foregoing, description has been made of the function of the encoding apparatus used for encoding the chronological PCM signals. Referring to the embodiment shown in <figref idref="f0018">Fig. 19</figref>, explanation is now given of the conversion from the A-mode encoded data into the B-mode encoded data and the of conversion from the B-mode encoded data into the A-mode encoded data.

0101First, as for the conversion from the A-mode to the B-mode, the encoded data A encoded with the A-mode is supplied to an input terminal 501, while the encoded orthogonal transform block size information is supplied to an input terminal 503. The orthogonal transform block size information is code-converted in a code converter 508 from the code indicating the orthogonal transform block size for the A-mode into the code indicating that for the B-mode. The code-converted information is transmitted to a bit allocation calculating circuit 507, while also being outputted at an output terminal 513.

0102The function of the code converter 508 is simply to group together the orthogonal transform block size information for two blocks for the A-mode and to encode the resulting information by the coding for the B-mode without causing any changes in the contents of the information.

0103The encoded data A is sent to an adaptive bit allocation and encoding circuit 505 where it is restored into spectral data by decoding and inverse quantization. The resulting spectral data is transmitted to a bit allocation and calculating circuit 507 for bit allocation. The bit allocation and calculating circuit 507 has the same function as that of the above-mentioned bit allocation calculating circuit 107.

0104The restored spectral data is sent to an adaptive bit allocation and encoding circuit 506 for the B-mode for encoding according to the B-mode as mentioned above. The quantized and encoded data B is outputted at an output terminal 511. Thus the conversion from the A-mode into the B-mode is executed by a simplified circuit consisting of the combination of the decoding circuit for the A-mode and the encoding circuit for the B-mode to enable high-speed conversion.

0105For conversion from the B-mode to the A-mode, the encoded data B (encoded with the B-mode) is supplied to an input terminal 512, while the encoded orthogonal transform block size information is supplied to an input terminal 514. The orthogonal transform block size information is code-converted by a code converter 509 from the code indicating the orthogonal transform block size for the B-mode into the code indicating the orthogonal transform block size for the A-mode. The code-converted block size information is transmitted to a format converting circuit 510, while being outputted at an output terminal 504.

0106The function of the code converter 509 is the reverse of that of the code converter 508 and consists simply in dividing the coded orthogonal transform block size information for the B-mode into coded data for two blocks for the A-mode.

0107The encoded data B is sent to the format converting circuit 510 whereby the encoded data is directly converted into data of the format of the A-mode before being outputted at an output terminal 502. There is no substantial change in the bit rate between the two modes such that only about one half of the main information is employed on the A-mode format. It is also possible to effect decoding and re-allocation of bits followed by encoding as in the above-mentioned conversion from the A-mode to the B-mode. However, the sound quality is deteriorated by re-quantization even although the substantial information volume is increased. Since the conversion from the B-mode into the A-mode may be achieved by format conversion, that is by simple re-arraying of coded data, the speed of conversion may further be raised.

0108Referring to <figref idref="f0019">Fig. 20</figref>, the decoding apparatus is explained. The encoded data in the frequency domain is supplied from the output terminal 110 of <figref idref="f0003">Fig. 4</figref> to an input terminal 210. The encoded data is first supplied to a decoding circuit 208 for adaptive bit allocation and thereby decoded into spectral data in the frequency domain.

0109The orthogonal transform block size information from the encoding apparatus of <figref idref="f0018">Fig. 19</figref> is supplied to an input terminal 211 and thence supplied to inverse orthogonal transform circuits 203, 204 and 205 for the respective bands. Of the above spectral data, the data for the band from 0 to 5.5 kHz is supplied to the inverse transform circuit 203 and the data for the band from 5.5 to 11 kHz is supplied to the inverse transform circuit 204, while the data for the band of from 11 to 22 kHz is supplied to the inverse transform circuit 205. By these circuits 203 to 205, the spectral data is inverse orthogonal transformed, depending on the above-mentioned orthogonal block size information.

0110The outputs of the inverse orthogonal transform circuits 204 and 205 are synthesized by a band-synthesizing filter 202, while the outputs of the inverse orthogonal transform circuit 203 and the synthesizing filter 202 are synthesized by a band-synthesizing filter 201 so as to be outputted at an output terminal 200 as replay signals.

0111The present invention is not limited to the above-described embodiments. For example, the two recording/replay media may be interconnected by a data transfer cable without it being necessary to have the two media built into one system. The present invention may be applied not only to audio PCM signals but also to digital speech signals or digital video signals. The synthesis of the above-mentioned minimum audibility curve may also be omitted, in which case the minimum audibility curve generating circuit 32 or the synthesizing circuit 27 may be omitted and an output of the subtractor 24 may be directly transmitted to the subtractor 28 after deconvolution by the divider 26. In addition, a wide variety of bit allocation techniques may be employed, such as fixed bit allocation, bit allocation based on signal band energies or bit allocation consisting of the combination of fixed bit allocation and variable bit allocation.

0112It is seen from the above that there is provided an apparatus and method for recording compressed data onto a recording medium therefor, whereby it becomes possible to prevent the sampling frequency signal generating circuit and hence the hardware from being increased in scale by employing the same sampling frequency despite the difference in the bit rate appropriate for different modes.

0113When encoding with the low bit rate, by using the same subsidiary information for plural block floating bands neighboring each other in the time domain or plural block floating bands neighboring each other along the frequency domain in one and the same time block for recording or transmission, and by changing the combination of the block floating bands for which the subsidiary information is to be used in common, such changes depending upon the orthogonal transform block size, it becomes possible to reduce the volume of the subsidiary information while minimizing the effect on the quantization and encoding efficiency. The subsidiary information thus saved may be allocated to the main information to improve the quantization and encoding efficiency.

0114If it is desired to convert high bit rate mode compressed signals into low bit rate mode signals for long time recording, the low bit rate compressed signals may be produced solely by data conversion in the frequency domain without the necessity of converting the original frequency-domain compressed signals into corresponding time-domain signals. High-speed signal transformation may be achieved because the process of orthogonal transform, inverse orthogonal transform and band-dividing and synthesizing filtering may be eliminated.