Variable rate coder.

0001The present application is divided from European Patent Application 89 117 738.8.

BACKGROUND OF THE INVENTION

(1) Field of the Invention

0002The present invention relates to a variable rate coder for transforming an input signal into a low and adaptively variable bit rate digital signal, comprising: a plurality of coding circuits each comprising an output quantizer having a different output bit rate, and independently transforming the input signal into a compressed digital signal.

0003Generally, when a voice signal or a picture signal is transmitted, the original input signal is coded at the sender side by a coding method to reduce the bit rate on the transmission line, and the coded signal is received and decoded to regenerate the original signal in the receiver side.

0004In the above transmission using coding, in addition to the requirement of a low bit rate, a sufficient quality of the regenerated signal is required according to the purpose of usage.

0005In particular, in communication systems which allow, variations of bit rate, such as a packet communication system, it is desirable to code an input signal to a least bit rate signal which maintains a sufficient quality of the regenerated signal from the view point of efficiency and quality.

(2) Description of the Related Art

0006One attempt to realize the above-mentioned coding wherein an input signal is coded to a least bit rate signal which maintains a sufficient quality of the regenerated signal, called a variable rate coder, is described by T. Taniguchi, et al. in "ADPCM with a Multiquantizer for speech Coding", IEEE Journal on Selected Areas in Communications, Vol. 6, Feb. 1988, pp.410-424.

0007The coder described in the above paper comprises a plurality of ADPCM coder units, and the number of bits of the quantizer output in each coder unit is different.

0008Input signals are coded in parallel in the plurality of coder units, the quality of the coding in each coder unit is evaluated for each frame of data, and the coder unit giving a lowest bit rate among the coder units satisfying a predetermined quality requirement is selected to be used for transmission.

0009To carry out the above evaluation (determination whether or not the output of each coder unit satisfies the above predetermined requirement of quality), two methods are described in the above paper.

0010In the one method, a signal-to-noise ratio (SNR) is obtained for each coder unit and each frame of data, and the obtained SNR value is compared with a threshold of the SNR corresponding to the above predetermined quality requirement.

0011In the coder carrying out a prediction coding, a difference between an input signal and a prediction value of the input are quantized by the quantizer, and the output of the quantizer is used as the output of the coder. The SNR in the coder is expressed by a multiple (summation when the SNR is expressed as a logarithm) of a prediction gain and a quantization gain when the input signal is large enough to ignore the quantization step size in the analog to digital converter which is provided in the input stage of the coder. Therefore, when the prediction gain is large (the above difference is small), only a small number of the quantization steps in the quantizer (a small number of output bits of the quantizer) is necessary to obtain the required SNR, and when the prediction gain is small (the above difference is large), a large number of the quantization steps in the quantizer (a large number of output bits of the quantizer) is necessary to obtain the required SNR. Thus, the above-mentioned evaluation of the SNR enables control of the output bit rate of the coder.

0012In the other method, a noise (error) level is used for the evaluation instead of the SNR.

0013As described in the above paper, there is a problem that a high bit rate output is selected for a low level input signal when evaluation based on the SNR is used, and for a high level input signal when evaluation based on the noise level is used.

0014Generally, an analog to digital converter is provided in the sampling stage of a coder for converting an analog input signal to a digital signal, where a fixed amplitude range for input signals (dynamic range) is assigned at the input of the analog to digital converter, and the step size (resolution) in the analog to digital converter is constant. The constant step size in the analog to digital converter limits the SNR in the analog to digital converter, and thus the SNR in the total coder output. In particular, when the input level becomes lower, it becomes difficult to obtain the required SNR value.

0015As mentioned above, the above control of the output bit rate of the coder using the evaluation of SNR is effective when the input signal is large enough to ignore the quantization step size in the analog to digital converter. However, when an SNR value in the analog to digital converter which is determined by the factor (input signal level)/(quantization step size in the analog to digital converter), is comparable with the SNR value required for the total coder (i. e., required for a signal regenerated in the receiver side through the coder and a corresponding decoder), the above control of the output bit rate of the coder using the evaluation of SNR is not effective because an increase in the quantization gain in the quantizer can recover only a decrease in the prediction gain, but cannot recover the SNR degraded in the analog to digital converter. Nevertheless, in the conventional variable rate coder, the above control of the output bit rate of the coder using the evaluation of SNR, i. e., an increase in the output bit rate of the coder when the SNR of the coder becomes lower, is carried out even when the SNR of the coder is degraded by the SNR in the analog to digital converter due to a low level input.

0016In the coder wherein the above control of the output bit rate is carried out based on the noise level instead of the SNR, the aforementioned problem of a high bit rate for a high level input is caused by a constant threshold level for noises which requires a high SNR for a high level input.

0017Therefore, in the conventional variable rate coders, even if the output bit rate of the coder is controlled based on the SNR or the noise level, there is a problem that a high bit rate which does not contribute to the quality of the transmitted signal is output according to the variation of the input level.

0018Further, in the conventional variable rate coder as described in the aforementioned paper by Taniguchi et al., a plurality of coder units are provided where the number of the coder units is determined by the extent of the variation of the necessary bit rates to obtain a required quality, and an input signal is coded in parallel in all the coding units. Therefore, there is a problem that the scale of the hardware is very large.

SUMMARY OF THE INVENTION

0019An object of the present invention is to provide a variable rate coder for realizing an optimal coding according to a variation of an input level from the viewpoint of quality and bit rate.

0020Another object of the present invention is to provide a variable rate coder for realizing an optimal coding according to a variation of an input level from the viewpoint of quality and bit rate, with a reduced hardware size.

0021According to the first aspect of the present invention, there is provided a coder for transforming an input signal into a low bit rate digital signal, comprising: a plurality of coding means, a plurality of output determining means, a control determining means, and an output selecting means. The plurality of coding means each comprises an output quantizer having a different number of output bits, and independently transforms the above input signal into a compressed digital signal. The plurality of output determining means each determines an output of one of the above plurality of coding means as an output of the (variable rate) coder (according to the present invention) by an individual determining method, in which method, the quality of a signal which is regenerated from an output of each coding unit is evaluated by an individual predetermined quality evaluating standard, and then an output of a coding means having a smallest number of output bits among outputs of coding means satisfying the above standard is determined as the output of the coder. The control determining means selects one of the above output determining means as a means to determine the output of the coder, according to the power of the above input signal. The output selecting means selects an output of one of the plurality of coding means as the output of the coder, according to the output of the output determining means which is selected by the control determining means.

0022According to the second aspect of the present invention, there is provided a coder for transforming an input signal into a low bit rate digital signal, comprising: a plurality of coding means, a plurality of decoding means, a plurality of error obtaining means, a plurality of S/N obtaining means, an S/N control output determining means, an error level control output determining means, a control determining means, and an output selecting means. The plurality of coding means each comprises an output quantizer having a different number of output bits, and independently transforms the above input signal into a compressed digital signal. The plurality of decoding means are provided corresponding to the above plurality of coding means. The plurality of error obtaining means are provided corresponding to the above plurality of decoding means, and each obtaining an error which is generated through a corresponding coding means and a corresponding decoding means, based on the above input signal and an output of the corresponding decoding means. The plurality of S/N obtaining means are provided corresponding to the above plurality of decoding means, and each obtains a signal-to-noise ratio based on the power of the above input signal and the power of an output of a corresponding error obtaining means. The S/N control output determining means determines an output of a coding means among the outputs of the above plurality of coding means, as an output of the coder, according to the outputs of the above plurality of S/N obtaining means. The error level control output determining means determines an output of a coding means among the outputs of the above plurality of coding means, as an output of the coder, based on the outputs of the above plurality of error obtaining means. The control determining means which determines as an output of the coder an output of a coding means determined by the above S/N control output determining means when the power of the input signal is greater than a first predetermined value, the above control determining means determines as the output of the coder an output of a coding means determined by the above error level control output determining means when the power of the input signal is not greater than a second predetermined value, and the above control determining means determines as an output of the coder an output of a coding means having a smaller number of output bits between the above output determined by the above S/N control output determining means and the above output determined by the above error level control output determining means, when the above power of the input signal is between the above first and second predetermined values. The output selecting means determines as an output of the coder an output of a coding means which is determined by the above control determining means.

0023According to the third aspect of the present invention, there is provided a coder for transforming an input signal into a low bit rate digital signal, comprising: a plurality of coding means, a plurality of prediction gain obtaining means, an bit rate calculating means, a bit rate setting means, and an output selecting means. The plurality of coding means in each of which a number of output bits can be variably set, and each of which can carry out a predictive coding independently. The plurality of prediction gain obtaining means each for obtaining a prediction gain in each of the above plurality of coding means. The bit rate calculating means for determining a plurality of successive numbers as the above numbers of output bits of the above plurality of coding means, based on a required value of the signal-to-noise ratio of the coder and the above prediction gains in the above plurality of coding means. The bit rate setting means for setting the above plurality of successive numbers in the above plurality of coding means as the above number of output bits, respectively. The output selecting means for selecting an output of one of the above plurality of coding means, as the output of the above coder, based on a predetermined standard.

BRIEF DESCRIPTION OF THE DRAWINGS

0024In the drawings: <ul id="ul0001" list-style="none"><li>Figure 1 shows an embodiment of the first aspect of the present invention;</li><li>Figure 2 shows an embodiment of the second aspect of the present invention;</li><li>Figure 3 shows an embodiment of the third aspect of the present invention;</li><li>Figure 4 shows the construction of the bit rate presetting circuit; and</li><li>Figure 5 shows the construction of the coder.</li></ul>

DESCRIPTION OF THE PREFERRED EMBODIMENTS

0025Before describing the preferred embodiment of the present invention, the basic principle of the first aspect of the present invention is explained below.

0026As described in the summary of the invention, according to the first aspect of the present invention, an input signal is coded in parallel in the plurality of coding means, and the output of one of the plurality of coding means which is selected by one of the plurality of output determining means, is used as the output of the (variable rate )coder. The above one of the output determining means is further selected by the control determining means according to the power of the input signal.

0027Thus, when the characteristic (particularly the applicability or suitability) of the quality control using the individual predetermined evaluation standard by each of the output determining means depends on the power of the input signal (or input signal level), most suitable or applicable one of the quality controls using the individual predetermined evaluation standards by the output determining means is selected by the control determining means according to the power of the input signal (or input signal level).

0028Figure 1 shows an embodiment of the above-mentioned first aspect of the present invention.

0029In Fig. 1, reference numerals 111, 121, and 131 each denote a coding unit, 112, 122, and 132 each denote a decoding unit, 61, 62, and 63 each denote an adder, 64, 65, 66, and 67 each denote a power calculator, 21, 22, and 23 each denote a divider, 4a denotes a control select circuit, 24 denotes an S/N evaluating control circuit, 3a denotes a noise evaluating control circuit, 51 denotes a selector, 52 denotes a selector and multiplexer, 7 denotes a transmission line, 8 denotes a demultiplexer, and 9 denotes a decoding unit.

0030Among the above composing elements, the demultiplexer 8 and decoding unit 9 are included in the receiver side, all the other composing elements except the above elements in the receiver side, and the transmission line 7, constitute a construction in the sender side.

0031A input signal (denoted by SIN in Fig. 1) is input in parallel into the coding units 111, 121, and 131.

0032The coding units 111, 121, and 131 realize the aforementioned plurality of coding means, have different numbers of output bits (for example, three, four, and five, respectively), and the coding units 111, 121, and 131 in Fig. 1 each carry out, for example, ADPCM coding. The output (coded) signals (each denoted by SCOD1, SCOD2, and SCOD3) from the coders 111, 121, and 131, are input into the selector and multiplexer 52.

0033The decoding units 112, 122, and 132 respectively decode the output signals (each denoted by SCOD1, SCOD2, and SCOD3) from the corresponding coders, and therefore, each of the decoding units 112, 122, and 132 regenerate the input signal suffering a possible degradation by the coding and decoding through the corresponding coding unit and decoding unit, respectively.

0034The outputs of the decoding units 112, 122, and 132 (the above-mentioned regenerated signals, and each denoted by SDEC1, SDEC2, and SDEC3) are each subtracted from the real input signal in the adder 61, 62, and 63, respectively, and thus, the error signals which are each generated through the coding units 111, 121, and 131, and the decoding units 112, 122, and 132, respectively, are obtained.

0035The error signals are each input into the power calculators 65, 66, and 67, respectively, and thus, the powers (each denoted by PWR1, PWR2, and PWR3) of the above error signals are obtained.

0036On the other hand, the power (denoted by PWR0) of the input signal is calculated in the power calculator 64, and, signal-to-noise ratios (each denoted by SN1, SN2, and SN3) for the pairs of the coding units and the corresponding decoding units 111 &amp; 112, 121 &amp; 122, and 131 &amp; 132, are respectively calculated in the dividers 21, 22, and 23.

0037The above signal-to-noise ratios SN1, SN2, and SN3 from the dividers 21, 22, and 23 are input into the S/N evaluating control circuit 24.

0038The S/N evaluating control circuit 24 corresponds to one of the aforementioned output determining means in the first aspect of the present invention, and determines an output of one of the coding units 111, 121, and 131, as an output of the coder (i. e., the output of the construction in the sender side as shown in Fig. 1), according to the above signal-to-noise ratios SN1, SN2, and SN3.

0039In detail, the S/N evaluating control circuit 24 first determines the outputs of the coding units corresponding to the outputs of the dividers each satisfying a predetermined S/N condition, and then determines an output of a coding unit having a smallest number of output bits among the coding units satisfying the predetermined S/N condition, as the output of the coder.

0040The above S/N condition is, for example, that the signal-to-noise ratio must be greater than a predetermined threshold. The threshold value (denoted by SNTH in Fig. 1) is predetermined to, for example, 36 dB.

0041Thus, if, for example, the signal-to-noise ratios SN2 and SN3 from the dividers 22 and 23 are above 36 dB, and the signal-to-noise ratio SN1 from the divider 21 is below 36 dB, first, it is determined that the coding units 121 and 131 satisfy the above S/N condition, and then the output of the coding unit 121 is determined as the output of the coder because the number of output bits (four) of the coding unit 121 is smaller than the number of output bits (five) of the coding unit 122, at the stage of the S/N evaluating control circuit 24. The result of the determination is output in the form of a selector control signal (denoted by ESN in Fig. 1), and is then applied to one input terminal of selector 51.

0042On the other hand, the powers of the error signals PWR1, PWR2, and PWR3 are input into the noise evaluating control circuit 3a.

0043The noise evaluating control circuit 3a corresponds to another of the aforementioned output determining means in the first aspect of the present invention, and determines an output of one of the coding units 111, 121, and 131, as an output of the coder (i.e., the output of the construction in the sender side as shown in Fig. 1), according to the above powers of the error signals PWR1, PWR2, and PWR3.

0044In detail, the noise evaluating control circuit 3a first determines the outputs of the coding units corresponding to the outputs of the dividers each satisfying a predetermined error condition, and then determines an output of a coding unit having a smallest number of output bits among the coding units satisfying the predetermined error condition, as the output of the coder.

0045The above error condition is, for example, that the power of the error signal must be smaller than a predetermined maximum. The maximum value (denoted by NTH in Fig. 1) is predetermined to, for example, -75 dBm0.

0046Thus, if, for example, the powers of the errors PWR2 and PWR3 from the power calculators 66 and 67 are below -75 dBm0, and the powers of the error PWR1 from the power calculator 65 is over -75 dBm0, first, it is determined that the coding units 121 and 131 satisfy the above error condition, and then the output of the coding unit 121 is determined as the output of the coder because the number of output bits (four) of the coding unit 121 is smaller than the number of output bits (five) of the coding unit 122, at the stage of the noise evaluating control circuit 3a. The result of the determination is output in the form of a selector control signal (denoted by EN in Fig. 1), and is then applied to another input terminal of selector 51.

0047The control select circuit 4a corresponds to the aforementioned control determining means in the first aspect of the present invention, and is realized by a comparator.

0048The comparator 4a compares the above-mentioned power PWR0 of the input signal with a threshold level (denoted by SPTH in Fig. 1) of the power of the input signal. The result of the comparison is output in the form of a selector control signal (denoted by ED in Fig. 1), and is then applied to a control input terminal of selector 51.

0049The above threshold level of the power of the input signal is set to, for example, -50dBm0.

0050Thus, when the power of the input signal is above the above threshold level SPTH, the control output ESN from the S/N evaluating control circuit 24 is selected in the selector 51. The output of the selector 51 (denoted by SEL) is applied to the selector and multiplexer 52 as a control signal for a selector portion of the selector and multiplexer 52. The selector and multiplexer 52 selects one of the output signals from the coding units 111, 121, and 131 according to the control signal supplied from the selector 51, and multiplexes the selected output of a coding unit and the above control signal to transmit them through the transmission line 7. Thus, the aforementioned control of the output bit rate of the coder using the evaluation of SNR is carried out for selecting a number of output bits in the coder (for selecting a coding unit).

0051When the power of the input signal is below the above threshold level SPTH, the control output EN from the noise evaluating control circuit 3a is selected in the selector 51. Similar to the above case, the output of the selector 51 (denoted by SEL) is applied to the selector and multiplexer 52 as a control signal for a selector portion of the selector and multiplexer 52. The selector and multiplexer 52 selects one of the output signals from the coding units 111, 121, and 131 according to the control signal supplied from the selector 51, and multiplexes the selected output of a coding unit and the above control signal to transmit them through the transmission line 7. Thus, the aforementioned control of the output bit rate of the coder using the evaluation of noise levels is carried out for selecting a number of output bits in the coder.

0052In the above multiplexing operation, coded data can be formed as a packet having a header portion and an information portion. In the header portion, the above control signal SEL which indicates which one of the outputs of the coding units are selected for the transmission, can be contained, and, in the information portion, for example, ADPCM codes (the above selected output) for 16 data samples are contained.

0053The decoder 9 shown in the receiver side, contains the same construction of three decoding units as decoding units 112, 122, and 132 in the sender side, and, when a frame of data is received in the receiver side, the above-mentioned control signal SEL which is transmitted in the header portion, is divided, and is supplied to the decoder 9 to select one of the above-mentioned decoding units corresponding to the selection in the sender side. Thus, the regenerated signal (denoted by REG) can be obtained in the receiver side.

0054In the above embodiment, all the calculations and controls in the above construction (and thus a switching for the selection of a bit rate) may be carried out for each frame.

0055Figure 2 shows an embodiment of the second aspect of the present invention.

0056The embodiment of the second aspect of the present invention is realized as a variation of the above embodiment of the first aspect of the present invention. The only difference in the constructions between the embodiments of the first and second aspects of the present invention is the function of the control select circuit. In Fig. 2, the control select circuit 4a' receives the control outputs ESN and EN from the S/N evaluating control circuit 24 and the noise evaluating control circuit 3a, and functions as follows.

0057As described in the summary of the invention, in the second aspect of the present invention, the range of the power of the input signal is divided into three regions by first and second predetermined threshold values (denoted by SPTH1 and SPTH2 in Fig. 2), where the first predetermined threshold value is greater than the second predetermined threshold value.

0058The control determining means (which can be realized by the control select circuit 4a' in Fig. 2) in the second aspect of the present invention, determines as an output of the coder an output of a coding unit determined by the S/N control output determining means (which can be realized by the S/N evaluating control circuit 24 in Fig. 2) when the power of the input signal is greater than a first predetermined value SPTH1, the above control determining means (the control select circuit 4a' in Fig. 2) determines as the output of the coder an output of a coding unit determined by the above error level control output determining means (which can be realized by the noise evaluating control circuit 3a in Fig. 2) when the power of the input signal is not greater than a second predetermined value SPTH2, and the above control determining means (the control select circuit 4a' in Fig. 2) determines as an output of the coder an output of a coding unit having a smaller number of output bits between the output determined by the S/N control output determining means (the control select circuit 4a' in Fig. 2) and the above output determined by the above error level control output determining means (the noise evaluating control circuit 3a in Fig. 2), when the above power of the input signal is between the above first and second predetermined values SPTH1 and SPTH2.

0059According to the second aspect of the present invention, a margin is given for determination of the threshold level of the power of the input signal. The first and second threshold levels SPTH1 and SPTH2 in the second aspect of the present invention may be set to ± 6 dB of the threshold level in the first aspect of the present invention.

0060In the first and second aspects of the present invention, the numbers of output bits (bit rate) in the plurality of coding means (units) in the variable rate coder, are each fixed, and therefore, the number of the coder units is determined by the extent of the variation of the necessary bit rates to obtain a required quality, and thus the scale of the hardware is very large.

0061As described in the summary of the invention, in the third aspect of the present invention, a number of output bits in each of the plurality of coding units (means) can be variably set, and the numbers of output bits which are to be set are calculated from a required value of the signal-to-noise ratio of the coder and the prediction gains in the plurality of coding units (in the third aspect of the present invention, each of the coding units is assumed to carry out a predictive coding). The calculated numbers of output bits are preset in the coding units before carrying out the aforementioned operations as the first and second aspects of the present invention.

0062Figure 3 shows an embodiment of the third aspect of the present invention.

0063In Fig. 3, basic functions of a plurality of coding units 101 to 10m, a plurality of decoding units 201 to 20m, a selector and multiplexer 500, a transmission line 7, a demultiplexer 600, and a decoder 601, are the same as the corresponding composing elements in the construction of Fig. 1 or 2, except that the numbers of output bits in the plurality of coding units can be variably set, and each of the coding units carries out a predictive coding.

0064In addition, the evaluation and control circuit 400 shown in Fig. 3 corresponds to the construction of the adders 61, 62, and 63, the power calculators 64, 65, 66, and 67, the dividers 21, 22, and 23, the control select circuit 4a or 4a', the S/N evaluating control circuit 24, the noise evaluating control circuit 3a, and the selector 51 shown in Fig. 1 or 2. As the characteristics of the third aspect of the present invention relate to the presetting of the number of output bits in the coding units 101 to 10m and the corresponding decoding units 201 to 20m, the evaluation and control circuit 400 can be either construction shown in Fig. 1 or 2.

0065The bit rate preset circuit 300 shown in Fig. 3 obtains a plurality of successive numbers for presetting the numbers of output bits of the plurality of coding units, by calculation using a required value of the signal-to-noise ratio (denoted by SN in Fig. 3) of the coder, a resolution (denoted by N_O in Fig. 3) of an analog to digital converter which is used for sampling the analog input signal, and prediction gains supplied from the coding units 101 to 10m.

0066The calculation carried out in the bit rate preset circuit 300 is as follows.

0067First, the center value i of the above plurality of successive numbers is calculated according to the equations, <maths id="math0001" num=""><math display="inline"><mrow><mtext>i=(SN-PG+10.8-20log₁₀AR)/6</mtext></mrow></math><img file="EP0631402A2_D0001.tif" /></maths> when SN ≦ SNI, and <maths id="math0002" num=""><math display="inline"><mrow><mtext>i=(SNI-PG+10.8-20log₁₀AR)/6</mtext></mrow></math><img file="EP0631402A2_D0002.tif" /></maths> when SN ≧ SNI, where SN is the required value of the signal-to-noise ratio of the coder, AR is an input amplitude range expressed by a root mean square of the input signal, PG is a prediction gain in a coding unit, which is defined as <maths id="math0003" num=""><math display="inline"><mrow><mtext>PG=10log₁₀(S/E)</mtext></mrow></math><img file="EP0631402A2_D0003.tif" /></maths> , SNI is an ADC signal-to-noise ratio in the above-mentioned analog to digital converter, which is defined as <maths id="math0004" num=""><math display="inline"><mrow><msub><mrow><mtext>SNI=10log₁₀(S/N</mtext></mrow><mrow><mtext>O</mtext></mrow></msub><mtext>)</mtext></mrow></math><img file="EP0631402A2_D0004.tif" /></maths> , S is the power of the input signal, and E is a prediction gain in a coding unit.

0068The above equation <maths id="math0005" num=""><math display="inline"><mrow><mtext>i=(SN-PG+10.8-20log₁₀AR)/6</mtext></mrow></math><img file="EP0631402A2_D0005.tif" /></maths> when SN ≦ SNI, is deduced from the fact that a summation of a prediction gain (logarithm) and a quantization gain (logarithm) contribute to the total signal-to-noise ratio, and the well known relation between the SNR and the number of output bits in a quantizer (for example, in N. S. Jayant, "Digital Coding of Speech Waveforms: PCM, DPCM, and DM Quantizers," Proceedings of the IEEE, Vol. 62, No. 5, PP. 611- , May 1974).

0069Since, when SN ≧ SNI, the signal-to-noise ratio in the analog to digital converter governs the total signal-to-noise ratio in the coder, SNI instead of SN is used in the above equation <maths id="math0006" num=""><math display="inline"><mrow><mtext>i=(SNI-PG+10.8-20log₁₀AR)/6</mtext></mrow></math><img file="EP0631402A2_D0006.tif" /></maths> .

0070Next, the plurality of successive numbers i-k to i+k are determined as the successive 2k+1 numbers having the above-obtained value i as the center value, where the value k is an integer, which is determined by the limitation on the hardware size, i. e., the limitation is determined by the maximum number of the coding units which can operate concurrently. For example, when i=3 and k=1, the plurality of successive numbers i-k to i+k are 2, 3, and 4.

0071Figure 4 shows a more detailed construction of the bit rate preset circuit 300 in Fig. 3.

0072In Fig. 4, reference numerals 310, 320 to 32m each denote a power calculator, 331 to 33m each denote a prediction gain calculating circuit, 341 to 34m, and 370 each denote a delay circuit, 360 denotes a bit rate calculating circuit, and 380 denotes a bit rate setting circuit.

0073The power calculator 310 calculates the power S of the input signal, and each of the power calculators 321 to 32m calculates the power E of a prediction error which is supplied from a corresponding coding unit as explained later.

0074The prediction gain calculating circuits 331 to 33m each calculate a prediction gain PG from the power S of the input signal and the corresponding prediction error E, as <maths id="math0007" num=""><math display="inline"><mrow><mtext>PG=S/E</mtext></mrow></math><img file="EP0631402A2_D0007.tif" /></maths> .

0075The calculations and presetting of the bit rates by the bit rate presetting circuit 300 is carried out for each frame of data. The prediction gains PG in the coding units are held in corresponding delay circuits 341 to 34m until the timing of a next frame change.

0076On the other hand, previously selected bit rates in the coding units 101 to 10m are held in the delay circuit 370.

0077At the timing of the frame change, the select control circuit 35 inputs all the values held in the delay circuits 331 to 33m, and 370, and then supplies them to the bit rate calculating circuit 360. The above-mentioned calculations to obtain the number of output bits (bit rate) in the coding units are carried out in the bit rate calculating circuit 360. Thus, the obtained values are preset in the corresponding coding units. In addition, the output of the bit rate calculating circuit 360 is also supplied to the evaluation control circuit 400, the information on the numbers of output bits is then supplied to the selector and multiplexer 52, and is multiplexed with the selected output of a coding unit to be transmitted through the transmission line 7.

0078Figure 5 shows the construction of a coding unit.

0079In Fig. 5, reference numeral 151 and 154 each denote an adder, 152 denotes a quantizer, 153 denotes an inverse quantizer, 155 denotes an adaptive predictor, and 156 denotes a scale factor register.

0080In the construction of Fig. 5, in the adder 151, a difference between an input signal SIN and an output of the predictor 155 is obtained, and the output of the adder 151 (prediction error PE) is quantized in the quantizer 152 to a preset bit rate for being transmitted as a coded signal COD. The prediction error PE is sent to a corresponding power calculator shown in Fig. 4.

0081The output COD of the quantizer 152 is also supplied to the inverse quantizer 153 to be inverse-quantized, and the output of the inverse quantizer 153 is added to the output of the predictor 155 at the adder 154. The output of the adder 154 is supplied to the predictor 155 to generate a next prediction data.

0082The variable setting of the number of output bits is carried out by changing a transformation table in the quantizer 152 and the corresponding inverse quantizer 153 (the transformation table determines the input-output relationship in the quantizer 152 and the corresponding inverse quantizer 153).

0083Since the construction of all the above embodiments are realized by digital signal processors, the above transformation tables for predetermined numbers of output bits (numbers of quantization steps) are installed in advance for each quantizer and a corresponding inverse quantizer. Thus, one of the transformation tables is selected by the output of the scale factor register 156. The content of the scale factor register 156 is renewed by the output of the aforementioned bit rate presetting circuit 300.

0084As explained above, according to the variable rate coder by the present invention, most suitable bit rate control is selected corresponding to variation of the input level (power), and the first and second aspects of the present invention are applicable to all types of coders. Further, according to the third aspect of the present invention, the numbers of output bits in the plurality of coding units are adaptively preset according to the prediction gains in the coding units, the power of the input signal, the resolution of the analog to digital converter, and a required SNR level. Thus, a relatively small number of coding units are necessary compared with the possible extent of the numbers of output bits of coding units. The third aspect of the present invention is applicable to all types of predictive coders wherein the number of output bits is variable.

0085Reference signs in the claims are intended for better understanding and shall not limit the scope.