Noise reduction apparatus

BAKCGROUD OF THE INVENTION

0001The present invention relates to a noise reduction apparatus at the time of an audio signal receiving, and more specifically to a noise reduction apparatus used for, for example, a car radio in which a pulsive noise is easily mixed, and to an audio output apparatus.

0002For example, when the electromagnetic noise in the circumstance of the car is considered, various pulsive electromagnetic noises (sometimes called the pulsive noise) such as an ignition noise, or mirror noise, are generated. Because these pulsive noises mix into a reception antenna connected to a car radio inside the car, it is ordinarily often experienced that the pulsive noise is generated in an output audio signal, therefore, generally for the car radio, the noise reduction apparatus to remove the pulsive noise is used.

0003Fig. 8 is a block diagram of the conventional (pulsive) noise reduction apparatus disclosed in, for example, JP-A-63-87026. In the drawing, when an FM intermediate frequency signal of an FM receiver is inputted, a detection signal outputted from an FM detection circuit 1 is supplied to a delay circuit 2 composed of a LPF (low pass filter) and delayed, and the output of the delay circuit 2 is supplied to a stereo demodulation circuit 5 through a gate circuit 3, and a level hold circuit 4. Further, the detection signal is supplied to a HPF (high pass filter) 6, and a noise component signal passed through the HPF 6 is amplified by a noise amplifier 7, and supplied to a noise detection circuit 8.

0004The noise detection circuit 8 is composed of a rectifier circuit to rectify the output signal of the noise amplifier 7, and a noise detection output is obtained thereby. This noise detection output is supplied to a waveform shaping circuit 9 and an integration circuit 10. Incidentally, a noise detection means 11 is structured by including the HPF 6, noise amplifier 7, noise detection circuit 8, waveform shaping circuit 9 and integration circuit 10.

0005The waveform shaping circuit 9 transforms the noise detection output into a pulse having a pulse width of a predetermined time period width, and supplied to the gate circuit 3. The gate circuit 3 is driven by a pulse supplied from the waveform shaping circuit 9 to the gate circuit 3 and comes to a signal cut out condition, and at the time of the signal cut out condition, a delay output level before the signal cut out is held by the level hold circuit 4, and supplied to the stereo demodulation circuit 5.

0006According to this, the generation of a spike noise by the sudden change of the potential of the demodulation signal due to the pulsive noise is prevented. When the pulse is not supplied from the waveform shaping circuit 9, the gate circuit 3 and the level hold circuit 4 become the signal through condition (through).

0007Further, the integration circuit 10 smoothes the noise detection output and obtains a DC signal corresponding to the noise level, and supplies the output of the integration circuit 10 to the noise amplifier 7 (feedback), and thereby, an AGC loop is formed.

0008Incidentally, the delay circuit 2 is provided to make up for a time period from the time when the pulsive noise is supplied to the HPF 6 to the time when the gate circuit 3 is made to be in the cut out condition. Further, because, in the stereo demodulation circuit 5, as shown in Fig. 9, the Lch (left channel) signal and Rch (right channel) signal are inputted in the form which is balancedly modulated with the frequency of 38 kHz around (Lch + Rch) /2, the LCh signal and the Rch signal can be separately picked out by the time division with, for example, 38 kHz.

0009Further, there is also a method by which the signal is corrected by an average value from the levels before and after the pulsive noise is generated, other than a method by which the former signal is level-held and outputted, as described above. Incidentally, in this method, the following problem exists.

0010In Fig. 10A, a waveform in the case where a correction error in which a low frequency signal is corrected to a correction period, becomes maximum, is shown. In the drawing, ○ mark is a value in which ● mark is corrected, and the difference between ○ mark and ● mark shows the correction error.

0011Next, in Fig. 10B, a waveform when a high freqency signal is corrected to a correction period, is shown. In the drawing, O mark shows a value in which ● mark is corrected. In the same manner as in Fig. 10A, the difference between ○ mark and ● mark shows the correction error.

0012Herein, when each of correction errors is observed, Fig. 10B is larger. That is, it is found that the relative relationship of the time width of the frequency to the correction period is very important, and even when a signal of a high frequency component is corrected, the error is large. Accordingly, even when the correction is conducted on the signal of the high frequency component, the correction error is heard as a noise. Herein, in contrast to that the pulse width of the pulsive noise is several tens µs to several hundreds µs, a composite signal has, as shown in Fig. 9, a component which is balancedly modulated with 38 kHz, and because the period of the signal is shorter than that of the pulsive noise, the correction error as shown in Fig. 10B, is generated.

0013The pulsive noise can be generated in a time period of ta of the demodulated signal in the FM demodulator 1. A value of A is held in a period of ta in the level hold circuit 4.

0014In this case because the Lch is held at the level of the Lch, the Lch after the stereo demodulation can be exactly corrected, however, the Rch is held at the level of the Lch, when the difference between the Rch and Lch is large, the large correction noise is generated.

0015EP-A-0 629 054 discloses a noise suppressor which operates by finding the sum and difference of two input signals, suppressing the noise in the generated difference signal and separating the signal in correspondence with the input signals, according to the difference and the sum signals.

0016US 4 574 390 discloses an impulse noise reduction circuit for FM receivers. A signal is derived from the outputs of the first and second sample-and-hold means which indicates the slope ratio of the demodulated signal at the moment immediately prior to the impulse noise, the slope ratio signal is applied to a linear integrator to generate a compensating voltage which is summed with the signal sampled by the first sample-and-hold means and maintained at the level which occured immediately prior to the occurence of the noise.

SUMMARY OF THE INVENTION

0017Aspects of the invention are provided in the accompanying claims.

0018It is desirable to obtain a noise reduction apparatus by which the correction error can be reduced even for the signal in which many high frequency components are included, and the noise suppression ability is increased.

0019A noise reduction apparatus accordig to a preferred embodiment of the present invention is provided with: a noise detection means for detecting a noise included in a demodulated audio signal; the first correction means for correcting the noise according to a signal value existing just before and just after a predetermined period including a generation time point of the noise in the demodulated audio signal which is detected by the noise detection means; the second correction means for correcting the noise according to a plurality of signal values respectively existing before and after the predetermined period including the generation time point of the noise in the demodulated audio signal which is detected by the noise detection means; a high band level detection means for detecting the level of a high band component of the audio signal; and a selection means for selecting either one of the first or the second correction means according to the output of the high band level detection means.

0020The noise reduction apparatus may be structured such that the first correction means outputs a low pass filter output of a signal value obtained from a linear interpolation of 2 signal values existing just before and just after a predetermined period including a generation time point of the noise.

0021The noise reduction apparatus may be structured such that the second correction means outputs a low pass filter output of the signal value obtained from the linear interpolation of 2 average signal values obtained by averaging a plurality of signal values existing before and after a predetermined period including the generation time point of the noise, corresponding to each of before and after the generation of the noise.

0022The noise reduction apparatus may be structured such that a level detection means for detecting the whole band level in the demodulated audio signal is further provided, and a selection means is operated according to a relationship between a ratio of the level output of a high band level detection means to the level output of the level detection means, -and a predetermined value.

0023The noise reduction apparatus may be structured such that the detection sensitivity of the noise detection means is changeable corresponding to the output level of the high band level detection means.

0024The noise reduction apparatus may be structured such that the selection means is operated according to the level of an addition signal and the level of a subtraction signal between the right channel signal and the left channel signal constituting the audio signal.

0025An audio output apparatus according to a further preferred embodiment of the present invention is provided with the noise reduction apparatus discussed above.

0026Further, it is desirable to obtain a noise reduction apparatus by which the correction error is not influenced by the other channel, and an audio apparatus.

BRIEF DESCRIPTION OF THE DRAWINGS

0027<ul id="ul0001" list-style="none" compact="compact"><li>Fig. 1 is a block diagram showing the structure of the embodiment 1.</li><li>Figs. 2A and 2B are views showing a corrected waveform of a noise reduction apparatus of the embodiment 1.</li><li>Figs. 3A and 3B are block diagrams showing the structure of a level detection means after the stereo demodulation of the noise reduction apparatus of the embodiment 1.</li><li>Fig. 4 is a block diagram showing the structure of the embodiment 2.</li><li>Figs. 5A and 5B are block diagrams explaining a correction operation of the embodiment 2.</li><li>Fig. 6 is a block diagram showing the structure of a noise detection means of the embodiment 2.</li><li>Fig. 7 is a block diagram showing the structure of the embodiment 3.</li><li>Fig. 8 is a block diagram showing the structure of the conventional noise reduction apparatus.</li><li>Fig. 9 is a view showing the FM stereo demodulation waveform.</li><li>Figs. 10A and 10B are an example of a noise correction waveform.</li></ul>

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

0028An embodiment having the structure by which a great effect can be displayed for the noise reduction, when being applied for, for example, a car audio device such as a car radio, an audio output apparatus such as a car video device of a car mounting type television, or an image sound apparatus including this audio output apparatus, will be described below.

0029Referring to the drawings showing the embodiment, the present invention will be specifically explained below.

Embodiment 1

0030Fig. 1 is a block structural diagram of a noise reduction apparatus of the first embodiment of the present invention.

0031Numeral 1000 is an FM demodulation means for demodulating an FM signal from a received broadcasting wave, numeral 5 is a stereo demodulation means, numeral 12 is an intermediate and low band correction means for intermediate and low band signals of an Rch of the stereo demodulation means 5, numeral 13 is a high band correction means for a high band signal of the Rch of the stereo demodulation means 5, numeral 21 is a switch for switching output signals of the high band correction means 13 and intermediate and low band correction means 12, and numeral 14 is an intermediate and low band correction means for intermediate and low band signals of an Lch of the stereo demodulation means 5 (herein, the intermediate and low band correction means 12 and 14 are the first correction means provided respectively corresponding to the Rch and Lch).

0032Numeral 15 is a high band correction means for the high band signal of the Lch of the stereo demodulation means 5, numeral 22 is a switch for switching the outputs of the high band correction means 15 and the intermediate and low band correction means 14, numeral 16 is a level detection means for detecting the level (envelope) of the output signal of the switch 21, and numeral 17 is a high band level detection means for detecting the high band component of the output signal of the switch 21, (herein, the high band correction means 13 and 15 are the second correction means provided respectively corresponding to the Rch and Lch).

0033Numeral 18 is a level detection means for detecting the level of the output signal of the switch 22, numeral 19 is a high band level detection means for detecting the high band component of the output signal of the switch 22, numeral 200 is a selection means for controlling the switch 21 corresponding to each output level of the level detection means 16 and the high band level detection means 17, and numeral 201 is a selection means for controlling the switch 22 corresponding to each output level of the level detection means 18 and the high band level detection means 19.

0034Next, operations will be described.

0035For example, in the car radio as an example of the above audio output apparatus, the broadcasting signal received by the attached antenna enters into the FM demodulator 1000, and the FM demodulation signal is outputted by the FM demodulator 1000. This FM demodulation signal is inputted into the respective stereo demodulation circuit 5 and noise detection means 110, and the following processing, which will be detailed below, is conducted.

0036Initially, the noise detection means 110 detects the pulsive noise, for example, in the same manner as in the noise detection means 11 in the conventional apparatus. As the output signal of the noise detection means 110, a gate signal of the high level (H level) is outputted, in the period in which the pulsive noise is detected, the gate signal of the low level (L level) is outputted, in the period in which the pulsive noise is not detected, and these gate signal outputs are inputted into the high band correction means 13, intermediate and low band correction means 12, the high band correction means 15, and intermediate and low band correction means 14.

0037Next, each of correction means 12 - 15 corrects the input signal (the output from the stereo demodulation means 5) in the period in which the gate signal is the H level, and the input signal is outputted as it is, in the period of the L level.

Correction by the values before and after the correction period

0038Herein, the intermediate and low band correction means 12 and 14 linear interpolate (the signal outputted by the linear interpolation is called the correction signal) the signal in the period in which the noise is generated, by using the values before and after the correction period. Incidentally, the correction signal is outputted through the low pass filter.

0039By using these intermediate and low band correction means 12 and 14, as the result that the signal whose wavelength is long (that is, the frequency is low) to the noise period, is linearly interpolated, an example of a waveform in a case where the correction error becomes the maximum, and as the result that the signal whose wavelength is short (that is, the frequency is high) to the noise period, is linearly interpolated, an example of a waveform in a case where the correction error becomes the maximum, are shown in Fig. 2.

0040In Fig. 2, ● mark shows the level to be originally obtained when the noise is not generated, and in this example, it corresponds to a point at which the correction error becomes the maximum, and ∇ mark shows the correction value (the correction value by the intermediate and low band correction means 12).

0041Fig. 2A shows a case where the wavelength of the output signal of the stereo demodulation means 5 is long to the correction period (that is, the frequency is low to the correction period), and the level difference (differential value)between ∇ mark and ● mark is small, the error by the correction is very small or slight to the amplitude of the signal waveform. As described above, in the first correction means, the correction signal for correcting the noise according to the signal values which exist just before and just after a predetermined period including the generation time point of the noise is outputted (in the first correction means in the description of each embodiment below, the same operation is conducted except for the specific exception).

0042Fig. 2B shows a case where the wavelength of the output signal of the stereo demodulation means 5 is short to the correction period (that is, the frequency is high to the correction period), and the level difference (differential value)between ∇ mark and ● mark is large, and the error by the correction is large to the amplitude of the signal waveform.

0043That is, when the above interpolation by using the intermediate and low band correction means 12 is conducted on the signal waveform in which the wavelength of the signal waveform is short (that is, the signal waveform of the high frequency), the satisfied suppression effect of the noise can not be obtained.

Correction by using the average value in the average period

0044Next, the high band correction means 13 and 15 conduct the averaging processing before and after the correction period, (◆ mark is the average value in the average period), and by using these two average values (the average signal value as the correction signal), the linear interpolation is conducted. Incidentally, this correction signal (average signal value) is outputted through the low pass filter.

0045Incidentally, the average period herein means a predetermined period before and after the noise period, and the average value of the signal level in the period is obtained according to a plurality of signal values included in the period.

0046By using the high band correction means 13 and 15, the waveforms in which the low frequency and the high frequency signals to the correction period are corrected, are shown in Fig. 2.

0047Fig. 2A shows a case where the wavelength of the output signal of the stereo demodulation means 5 is long to the correction period (that is, the frequency is low to the correction period), and in the level difference from ● mark, ∇ mark is smaller than ○ mark.

0048Fig. 2A shows a case where the wavelength of the output signal of the stereo demodulation means 5 is short to the correction period (that is, the frequency is high to the correction period), and in the level difference from ● mark, ○ mark is smaller than ∇ mark.

0049Accordingly, when the wavelength of the signal waveform is sufficiently long to the correction period (that is, the frequency of the signal waveform is low to the correction period), the correction (interpolation processing) is conducted by using the intermediate and low band correction means 12 and 14, and when the wavelength of the signal waveform is short to the correction period (that is, the frequency of the signal waveform is high to the correction period), the correction (interpolation processing) is conducted by using the high band correction means 13 and 15. As described above, in the second correction means, the correction signal for correcting the noise according to a plurality of signal values which exist just after a predetermined period including the generation time point of the noise is outputted (in the second correction means in the description of each embodiment below, the same operation is conducted except for the specific exception).

Level detection means

0050Next, the level detection means will be described (hereinafter, for easy understanding, initially the structure according to the series of the Rch will be described).

0051In the level detection means 16, the level of the signal corrected by using the high band correction means 12 or the intermediate and low band correction means 13, is detected (envelope detection).

0052The level detection means 16 in this case can be realized by adopting the structure, for example, as in Fig. 3A. Incidentally, herein, the DC component is not included in the output of the switch 21.

0053In the drawing, numeral 23 is an absolute value circuit, and numeral 24 is a low pass filter (LPF). Initially, in the absolute value circuit 23, the absolute value of the output signal outputted by the switch 21 is obtained, and the high band component is removed by the LPF 24. This output signal of the LPF 24 is outputted as the envelope of the signal outputted from the switch 21.

0054Incidentally, relating also to the series of the Lch, the level detector 18, for the high band correction means 15 or the intermediate and low band correction means 14 and the switch 22, the structure to which the Rch corresponds, is respectively adopted, and for also the structure of the level detector 18, the same structure as shown in Fig. 3A is adopted, and its operation is also the same.

High band level detection means

0055Next, the high band level detection means will be described (hereinafter, for simple understanding, initially, the structure according to the series of Rch will be described) .

0056The high band level detection means 17 detects the level of the signal which is corrected by using the high band correction means 12 and intermediate and low band correction means 13 (envelope detection).

0057The high band level detection means 17 in this case, can be realized by adopting the structure, for example, as shown in Fig. 3B. Incidentally, herein, a DC component is not included in the output of the switch 21.

0058In the drawing, numeral 25 is a high pass filter (HPF), numeral 26 is an absolute value circuit, and numeral 27 is a low pass filter (LPF) . Initially, in the HPF 25, the low band component of the output signal outputted from the switch 21 is removed, and the high band component is obtained.

0059Next, in the absolute value circuit 26, the absolute value of the output signal of the HPF 25 is obtained. Next, the high band component is removed by the LPF 27. The output signal of the LPF 27 is outputted as an envelope of the high band component of the signal outputted from the switch 21.

0060Incidentally, relating also to the series of the Lch, for the high band level detection means 19, the high band correction means 15 or the intermediate and low band correction means 14 and the switch 22, the structure to which the Rch corresponds, is respectively adopted, and for also the structure of the high band level detector 19, the same structure as shown in Fig. 3B is adopted, and its operation is also the same.

Selection means

0061Next, a selection means 200 will be described. Into the selection means 200, the output signal VH form the high band level detection means 17 and the output signal VA from the level detection means 16 are inputted.

0062Herein, when the VH/VA is smaller than a predetermined value (that is, the rate of the signal of the high band component is small), because it is considered that the rate of generation of the correction error generated for correcting the high band component signal is small, the selection means 200 connect the output side of the Rch to the intermediate and low band correction means 12 by the switch 21, or the selection means 201 connects the output side of the Lch to the intermediate and low band correction means 14 by the switch 22.

0063When the VH/VA is larger than a predetermined value (that is, the rate of the signal of the high band component is large), because it is considered that the rate of generation of the correction error generated for correcting the high band component signal is large, the selection means 200 connect the output side of the Rch to the high band correction means 13 by the switch 21, or the selection means 201 connects the output side of the Lch to the high band correction means 15 by the switch 22.

0064As described above, because the correction means is selected corresponding to the result of the comparison of the ratio of the level VH of the high band component (the envelope of the high band component) in the FM stereo demodulation signal to the level VA of the whole band (the envelope of the whole band), with the predetermined value, the correction error can be decreased.

0065Further, the above described these processing may also be conducted by the digital signal processing by using the DSP (Digital Signal Processor) after the output signal of the FM detection circuit 1000 is A/D converted (Analog to Digital conversion). In this case, in the correction means which is not selected in the intermediate and low band correction means 12 and 14, and the high band correction means 13 and 15, the processing for the correction may be neglected.

0066Further, the HPF 25 shown in Fig. 3 may be used as the high band level detection means 17 or 19 because, in the stereo demodulated signal, for example, the component higher than 15 kHz is basically unused. Therefore the BPF by which the component higher than 15 kHz can be removed, may be used.

0067Further, a case where the linear interpolation is used for the correction method, the signal in the noise period is linearly interpolated. As the signal passes through the LPF, and after the high band component of the correction error is suppressed, it may be replaced with the signal (noise) in the noise period.

0068Further, as a more simple type of the high band correction means 13 and 15, an input signal of the high band correction means 13 and 15 is input into an LPF, and holds an output signal level of LPF immediately before a pulse noise generation so as to correct.

0069Incidentally, in the above description, the operation of the selection means is determined according to the relationship between the rate (VH/VA) of the VH (the level output of the high band level detection means) to the VA (the level output of the level detection means), and a predetermined value, however, when the signal level of the VH is not extremely large, it is needless to explain that the operation of the selection means may be determined according to the relationship between only the VH and the predetermined value.

Embodiment 2

0070Fig. 4 is a block structural diagram of the noise reduction apparatus of the second embodiment of the present invention. Numeral 5 is a stereo demodulation means, numeral 12 is an intermediate and low band correction means for conducting the correction on the intermediate and low band signal of the Rch of the stereo demodulation means 5, and numeral 13 is a high band correction means for conducting the correction on the high band signal of the Rch of the stereo demodulation means 5.

0071Numeral 21 is a switch to switch the output signals of the high band correction means 13 and intermediate and low band correction means 12, numeral 14 is an intermediate and low band correction means for an intermediate and low band signal of the Lch of the stereo demodulation means 5, and numeral 15 is a high band correction means for a high band signal of the Lch of the stereo demodulation means 5.

0072Numeral 22 is a switch to switch the output signals of the high band correction means 15 and intermediate and low band correction means 14, numeral 16 is a level detection means of the output signal of the switch 21, and numeral 17 is a high band level detection means for detecting the high band component of the output signal of the switch 21.

0073Numeral 18 is a level detection means for detecting the level of the output signal of the switch 22, and numeral 19 is a high band level detection means for detecting the high band component of the output signal of the switch 22.

0074Numeral 28 is a selection means for controlling the switches 21 and 22 corresponding to each output level of the high band level detection means 17 and 19, and the level detection means 16 and 18, and numeral 111 is a noise-detection means for adjusting the sensitivity of the noise detection corresponding to the output from the high band level detection means 17 and 19.

0075Next, portions whose operation is different from the embodiment 1, will be described. In Fig. 5, a case where a small pulsive noise is corrected, is shown. (Herein, a noise having the amplitude up to 50 % of the amplitude level of the signal waveform is called a small pulsive noise).

0076In Fig. 5, Fig. 5A shows an example of a waveform before the small pulsive noise is corrected, and Fig. 5B shows an example of a waveform after the small pulsive noise is corrected. As can be seen from the comparison of Fig. 5A with Fig. 5B, in the example as shown in the drawing, from the result that the level difference (error) is larger in the waveform after the correction than in the waveform before correction (the waveform largely deforms from the original signal waveform. In the example shown in Fig. 5, the waveform is largely deformed in the portion which is corrected from the original sinusoidal wave), there is a case where the noise is rather larger by the correction. Specifically, when the high frequency signal is corrected, because the correction error becomes large, this tendency becomes large.

0077Accordingly, when the high frequency component is large, the detection sensitivity of the pulsive noise is lowered, so that the small noise is not detected, and the correction by the correction means 12 -15 is not conducted.

0078An example of the detection means 111 by which the above operations can re realized, is shown in Fig. 6. The operations of the HPF 6, noise amplifier 7, waveform shaping circuit 9, and integration circuit 10, which are shown in Fig. 6, are the same as the operations in the conventional apparatus. Further, in the adder 28, the output of the integration circuit 10 and the outputs of the high band level detection means 17 and 19, are given weight through a weight section 29 and a weight section 30 (a coefficient is respectively multiplied, of course, a case where the coefficient is 1, is included), and after that, each output is added, and the addition result is inputted into the noise amplifier 7 as the control signal.

0079Herein, the larger the above control signal (the result of addition) is, the smaller the noise amplifier 7 makes the gain. Accordingly, the gain of the noise amplifier 7 in the case where the outputs of the high band level detection means 17 and 19 are 0, acts so that the average level of the output signal of the noise detection circuit 8 is kept constant.

0080The average level of the output signal of the noise detection circuit 8 is smaller than the threshold value of the waveform shaping circuit 9. On the other hand, because, for the signal whose change in time is rapid, the gain of the noise amplifier 7 does not change, when the pulsive noise is added to the noise amplifier 7, the output of the noise-detection circuit 8 is larger than the threshold value of the waveform shaping circuit 9, and the waveform shaping circuit 9 outputs the H level and the pulsive noise is detected.

0081Herein, even when the pulsive noise smaller than the difference between the average value of the output of the noise detection circuit 8 and the threshold value of the waveform shaping circuit 9 is generated, it is not detected. Accordingly, when up to the small pulsive noise is detected, the difference between the average value of the output of the noise detection circuit 8 and the threshold value of the waveform shaping circuit 9 is made small, and when the small pulsive noise is not detected, the difference between the average value of the output of the noise detection circuit 8 and the threshold value of the waveform shaping circuit 9 may be made large.

0082Next, when many high band signals are included in the stereo demodulation signal, and outputs of the high band level detection means 17 and 19 become large, because the control signal of the noise amplifier becomes large, the gain of the noise amplifier 7 becomes small corresponding to that.

0083Accordingly, the average value of the output signal of the noise detection circuit 8 becomes small, and because the difference to the threshold value of the waveform shaping circuit 9 becomes large, the small pulsive noise is not detected.

0084As described above, when the level of the high band component of the FM demodulated signal of the stereo is large, because the detection sensitivity of the pulsive noise is lowered (that is, the detection sensitivity is changeable corresponding to the output level of the high band level detection means), the correction error due to the correction of the small pulsive noise is decreased.

0085Further, above described these processing may A/D convert the output signal of the FM detection circuit 1000, and the processing after that may be conducted by using the digital signal processing technique such as the DSP. Incidentally, in this case, in the intermediate and low band correction means 12 and 14, and the high band correction means 13 and 15, the processing for the correction in the correction means which is not selected, can be neglected.

Embodiment 3

0086Fig. 7 is a block structural diagram of the noise reduction apparatus of the third embodiment of the present invention. In the drawing, numeral 112 is a noise detection means for detecting the pulsive noise from the output of the FM detection circuit 1, numeral 120 is an intermediate and low band correction means for conducting the correction on the intermediate and low band signals when there are many intermediate and low band components, in the output signals of the FM detection circuit 1, numeral 130 is a high band correction means for conducting the correction on the high band signals when there are many high band components, in the output signals of the FM detection circuit 1, and numeral 21 is a switch to switch the outputs of the intermediate and low band correction means 120 and the high band correction means 130.

0087Numeral 5 is a stereo demodulation means connected to the output signal of the switch 21, numeral 160 is a level detection means for detecting the level of the Lch signal of the stereo demodulation means 5, numeral 170 is a high band level detection means for detecting the high band signal level in the Lch signals of the stereo demodulation means 5, and numeral 180 is a level detection means for detecting the signal level of the Rch signal of the stereo demodulation means 5.

0088Numeral 190 is a high band level detection means for detecting the high band signal level in the Rch signal of the stereo demodulation means 5, numeral 300 is an L - R level detection means for detecting the level of the L - R component which is the difference between the signal level of the Lch signal and the signal level of the Rch signal (the level of the subtraction signal. incidentally, R - L component may be allowable), and numeral 301 is an L + R level detection means for detecting the level of the L + R component which is the sum of the signal level of the Lch signal and the signal level of the Rch signal (the level of the addition signal).

0089Numeral 400 is a selection means for switching the switch 21 corresponding to each of outputs of the level detection means 160 and 180, high band detection means 170 and 190, and L - R level detection means 300.

0090Next, operations will be described. Initially, the noise detection means 112 detects the pulsive noise, for example, in the same manner as the detection means 11 in the conventional apparatus. The output signal of the noise detection means 112 outputs the high level (H level) in the period in which the pulsive noise is detected, and the low level (L level) in the period in which the pulsive noise is not detected, and it is inputted into the high band correction means 130 and the intermediate and low band correction means 120.

0091Next, the correction means 120 and 130 correct the signals in the period in which the gate signal is the H level. Herein, the output signal to be corrected of the FM detection circuit 1 is composed of the L + R component of 0 to 15 kHz, L - R component which is AM modulated with 38 kHz in the band of 23 to 53 kHz, and the pilot signal of 19 kHz.

0092Accordingly, when, on the signal in which many L - R components are included, the correction corresponding to the pulsive noise with the width of several tens µs is conducted, there is also a case in which the pulsive noises of several wavelengths exist in the correction period, and when the pre-value holding, or linear interpolation is simply conducted, there is a case in which the correction error rather becomes large. In this case, the high band correction means 130 is used, and the error due to the correction is made small.

0093Further, when the L - R components are small in the output signal of the FM detection circuit 1, the components of 23 kHz to 53 kHz are small, and when the high band components in the L + R components are small, because it is equivalent to that the high band components in the output signals of the FM detection circuit 1 are small, by simply conducting the pre-value holding or linear interpolation, the correction error can be made small.

0094Accordingly, in this case, when the following (1) and (2) are satisfied, the selection means 400 operates in such a manner that the switch 21 is connected to the intermediate and low band correction means 120. <ul id="ul0002" list-style="none" compact="compact"><li>(1) When the output of the L - R level correction means 300 is sufficiently smaller than the output signal of the L + R level detection means 301.</li><li>(2) When the output of the high band level detection means 170 and 190 is sufficiently smaller than the output of the level detection means 160 and 180.</li></ul>

0095Herein, the output signal of the L - R level detection means 300 is obtained from the output in which , for example, the absolute value of the difference between the stereo demodulated Lch and Rch is inputted into the LPF. Further, the output signal of the L + R level detection means 301 is obtained from the output of the LPF in which , for example, the absolute value of the sum of the Lch signal and Rch signal which are FM demodulated into the stereo, is an input.

0096As described above, when the above conditions of (1) and (2) are satisfied, because the high band signal components are few in the output signal from the FM detection circuit 1, when the correction is conducted simply by the linear interpolation, the difference from the original demodulation signal can be made rather small.

0097Further, the above described these processing may be conducted as follows: the output signal of the FM detection circuit 1 is A/D converted, and the subsequent processing is conducted by using the digital signal processing technique such as the DSP.

0098Further, in the description of the above embodiments, the processing is conducted on the signal after the stereo FM demodulation, and the processed signal is inputted into the selection means 400, however, the signal level of the high band in the composite signal in which the output signal of the switch 21 is corrected, is detected, and when it is small, because the L - R component is small, and further, the high band component of the signal which is FM demodulated into the stereo is also small, the switch 21 may be connected to the intermediate and low band correction means 120.

0099As described above, according to preferred embodiments of the present invention, the following effects can be obtained.

0100In the noise reduction apparatus according to a preferred embodiment of the present invention, the apparatus is provided with: a noise detection means for detecting a noise included in a demodulated audio signal; the first correction means for outputting a correction signal for correcting the noise according to a signal value existing just before and just after a predetermined period including a generation time point of the noise in the demodulated audio signal which is detected by the noise detection means; the second correction means for outputting the correction signal for correcting the noise according to a plurality of signal values respectively existing before and after the predetermined period including the generation time point of the noise in the demodulated audio signal which is detected by the noise detection means; a high band level detection means for detecting the level of a high band component of the audio signal; and a selection means for selecting either one of the first or the second correction means according to the output of the high band level detection means, thereby, even when the high frequency components are included in the audio signal, the high frequency components are detected, and when the rate of the high frequency components is large, because the correction in which the error is few to the high frequency signal, is selected, the correction error when the rate of the high frequency components is large, can be decreased.

0101In the noise reduction apparatus according to a further preferred embodiment of the present invention, because the first correction means is structured such that it outputs a low pass filter output of a signal value obtained from a linear interpolation of 2 signal values existing just before and just after a predetermined period including a generation time point of the noise, as a correction signal, thereby, the correction error when the rate of the low frequency components is large, can be decreased.

0102In the noise reduction apparatus according to a further preferred embodiment of the present invention, because the second correction means is structured such that it outputs a low pass filter output of the signal value obtained from the linear interpolation of 2 average signal values obtained by averaging a plurality of signal values existing before and after a predetermined period including the generation time point of the noise, corresponding to each of before and after of the generation of the noise, as the correction signal, thereby, the signal correction when the rate of the high frequency components is large, can be accurately corrected, and the correction error can be decreased.

0103In the noise reduction apparatus according to a further preferred embodiment of the present invention, because the apparatus is structured such that a level detection means for detecting the whole band in the demodulated audio signal is further provided, and a selection means is operated according to a relationship between a ratio of the level output of a high band level detection means to the level output of the level detection means, and a predetermined value, thereby, even when the output from the high band level detection means is large, the noise can be surely caught.

0104In the noise reduction apparatus according to a further preferred embodiment of the present invention, because the noise detection means is structured such that the detection sensitivity of the noise detection means is changeable corresponding to the output level of the high band level detection means, thereby, the generation of large error due to the correction in the case where the low level noise is included, can be prevented.

0105In the noise reduction apparatus according to a further preferred embodiment of the present invention, because the apparatus is structured such that the selection means is operated according to the level of an addition signal and the level of a subtraction signal between the right channel signal and the left channel signal constituting the audio signal, thereby, the correction suited to the received signal can be conducted.

0106In an audio output apparatus according to a further preferred embodiment of the present invention, because the apparatus is provided with the noise reduction apparatus described above, the audio output apparatus by which, even when the noise is included, the optimum correction to the noise is conducted, and the high quality audio output can be obtained, can be realized.