JP2012225982A - Speech processing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To attain soft mute for reducing abnormal noise to be generated due to deterioration of a reception state with a speech processing device which handles a digital signal.SOLUTION: A speech processing device includes: fast Fourier transform means 1 for performing Fourier transform to a digital input signal in a voice band at high speed; absolute value calculation means 2 for calculating an absolute value of the signal to which the Fourier transform is performed to generate an absolute value signal; auto-correlation function calculation means 3 for calculating an auto-correlation function of the calculated absolute value signal; difference calculation means 8 for calculating variation of output of the auto-correlation function calculation means 3; noise determination means 9 for calculating a noise component by the size of difference; fade-out coefficient derivation means 10 for calculating a fade-out amount based on a calculation result by the noise determination means 9; and volume control means 11 for controlling a level of the digital input signal based on a calculation result by the fade-out coefficient derivation means 10 to be output.

Description

  The present invention relates to an audio processing apparatus suitable for, for example, a wireless audio conference system using infrared rays and the like, and in particular, in a receiver that handles digital signals, it fades out without a sense of incongruity when the reception state deteriorates. It is a thing.

  In a wireless audio conference system, an audio signal captured by a microphone is transmitted using infrared rays or radio waves, this transmission signal is received by a receiver, restored to an audio signal, and emitted from a speaker as audio. Yes. Not only in wireless audio conferencing systems, but in systems that wirelessly transmit audio signals, receive them, and return them to audio signals, if the transmission and reception conditions are poor and the reception status deteriorates, the proportion of noise mixed in the audio increases. Discomfort to those who listen to audio.

  In particular, in the FM transmission / reception system, when the reception state deteriorates, white noise such as “Zar” occurs, and when the reception state further deteriorates, suddenly an unpleasant sound such as “pop” is emitted and the sound stops. End up. In addition, when the reception state improves, a voice is suddenly emitted and surprises the listener.

  In view of this, a so-called soft mute circuit has been proposed in which the level of the demodulated audio signal is continuously attenuated when the receiver generates an unpleasant noise due to the deterioration of the reception state. One of the soft mute circuits of the receiver described in Patent Document 1 is a level control circuit that controls the level of a demodulated audio signal, and an intermediate frequency signal that is generated by mixing an RF signal and a local oscillation signal. And a modulation degree detection circuit for generating a second control signal corresponding to the output level of the audio signal, wherein the first control signal is predetermined. The level control circuit is configured to control the output level of the audio signal in accordance with the output of the second control signal when the value becomes equal to or less than the value of the electric field strength.

  According to the invention described in Patent Document 1, the demodulation noise that increases when the electric field strength is reduced is reduced, and the attenuation characteristic of the output level is changed according to the degree of the modulation degree. There is no effect. However, the signal handled in the invention described in Patent Document 1 is an analog signal. In a receiver that handles digital signals, when noise is likely to occur due to deterioration of the reception state, so-called soft mute cannot be realized with the same technical idea as the invention described in Patent Document 1, and Patent Document 1 A completely different idea from the described invention is required.

  In the conventional microphone system using infrared rays, when the reception state deteriorates, white noise of an amount corresponding to the degree of deterioration is generated. Therefore, for example, when the energy of a specified frequency (for example, 50 kHz) other than the audible band generated from the FM demodulation circuit for analog signals exceeds a threshold value, muting is performed (for example, the output of the microphone is turned off). There is.

  However, in the above conventional microphone system, when the reception state is unstable and the energy of the specified frequency fluctuates around the threshold value, the microphone output is repeatedly turned on and off, and each time a difference such as “X, X” is given. Sound is generated. Further, when the FM demodulating circuit is constituted by a digital processing circuit, it is difficult to mute with the energy of the designated frequency other than the audible band.

Japanese Patent Publication No. 6-26316

  An object of the present invention is to realize so-called soft mute for reducing abnormal noise caused by deterioration of the reception state in a sound processing apparatus that handles digital signals.

The speech processing apparatus according to the present invention
A fast Fourier transform means for performing a Fourier transform of a digital input signal in a voice band at a high speed;
Absolute value calculating means for generating an absolute value signal by calculating an absolute value of the Fourier transformed signal;
Autocorrelation function calculating means for calculating an autocorrelation function of the calculated absolute value signal;
Difference calculating means for calculating the amount of change in output of the autocorrelation function calculating means;
Noise determining means for calculating a noise component according to the magnitude of the difference;
A fade-out coefficient deriving unit that calculates a fade-out amount based on the calculation result by the noise determination unit;
Volume adjusting means for adjusting and outputting the level of the digital input signal based on the calculation result by the fade-out coefficient deriving means;
It has the most important feature.

  According to the present invention, the reception state is detected using the autocorrelation method, the noise generated as the reception state deteriorates is quantified, and the level of the digital input signal is adjusted in accordance with the noise component to be faded out. Therefore, so-called soft mute can be realized by a digital audio signal circuit.

It is a functional block diagram which shows one Example of the audio processing apparatus which concerns on this invention. It is a functional block diagram which shows another Example of the audio processing apparatus which concerns on this invention. It is a wave form diagram which compares and shows the correlation function waveform of a signal with periodicity, and a non-periodic signal. It is a graph which shows the level of the noise amount with respect to a reception level, fade-out amount, and an output volume in percentage.

  Embodiments of a speech processing apparatus according to the present invention will be described below with reference to the drawings.

  The audio processing apparatus shown in FIG. 1 is, for example, a part of a general broadcast receiver, a communication receiver, or a wireless audio conference system using infrared rays or radio waves, and is demodulated. This is a voice processing apparatus. In FIG. 1, “input” refers to an input of a voice band signal, and here is input as a digital signal. The transmission / reception system is not limited to a digital system, and the transmission / reception system may be an analog system, and a demodulated signal may be converted into a digital signal, and the converted digital signal may be used as the “input”.

  The input signal passes through the fast Fourier transform means 1, the absolute value calculation means 2, the autocorrelation function calculation means 3, the difference calculation means 8, the noise determination means 9, the fade-out coefficient derivation means 10, and the volume control means 11 in this order. Predetermined processing is performed by each means. The processing performed by each of the above means is as follows.

  The fast Fourier transform means 1 performs fast Fourier transform (FFT) on a digital signal in an input audio band, and converts a function having a variable time and space coordinates into a function having a variable frequency. The input digital signal in the voice band is a time domain signal, and this input signal is converted into a frequency domain signal by the fast Fourier transform means 1.

  The digital voice band signal converted into the frequency domain signal by the fast Fourier transform means 1 is represented by a complex number. In order to calculate the gain of this signal, the absolute value calculation means 2 calculates the absolute value signal of the converted signal. Is generated.

  The autocorrelation function calculation means 3 calculates an autocorrelation function of the absolute value signal calculated by the absolute value calculation circuit 2. Autocorrelation is a measure for how well a signal matches itself with a time-shifted signal, and is expressed as a function of the magnitude of the time shift, and this function is called an autocorrelation function. If the amplitude of a signal of a certain length in time and a signal delayed in time are large and there are similar repetitive components, the correlation value or autocorrelation function of the two signals becomes large. Therefore, since a correlation value of a noise signal including various sounds is immediately smaller than that of an audio signal such as music, the autocorrelation function of the noise signal is smaller than the correlation function of the audio signal. The autocorrelation function calculation means 3 performs fast Fourier inverse transform in order to calculate the autocorrelation function of the absolute value signal, and returns it to the time domain signal.

  The difference calculation means 8 performs high-pass filter processing in order to obtain the difference, that is, the amount of change in the amplitude value of the correlation waveform included in the autocorrelation function obtained by the autocorrelation function calculation means 3. When this high-pass filter processing is performed, that is, when the correlation waveform is passed through a high-pass filter, noise, for example, non-periodic signals such as white noise and pink noise have a large difference, that is, a change amount. Become. On the other hand, when a sound signal such as music having a large number of periodic signal components is passed through a high-pass filter, it becomes a gentle waveform and does not quickly become small, and is durable. FIG. 3 shows a comparison of correlation function waveforms of a periodic signal and an aperiodic noise signal. In FIG. 3, a waveform A shows a correlation function waveform when there are many periodic sound signal components, and a waveform B shows a correlation function waveform when there are many non-periodic noise signal components.

  As can be seen from FIG. 3, it is possible to determine whether or not the signal has a large number of non-periodic noise signal components from the difference in the steepness of the correlation function waveform, that is, the difference in change amount. Therefore, when the difference is large from the difference value obtained by the difference calculation unit 8, the noise determination unit 9 determines that there are many non-periodic noise signal components because the inclination of the correlation function waveform is large. On the other hand, when the difference is small, it is determined that the non-periodic noise signal component is small, that is, the periodic audio signal component is large, because the slope of the correlation function waveform is small. A signal representing the determination result is output to the next fade-out coefficient deriving means 10. In order to finely adjust the volume, which will be described later, it is desirable that the noise determination in the noise determination means 9 is performed in multiple stages. However, it may be determined by setting a certain threshold and determining whether or not the threshold is exceeded.

  The fade-out coefficient deriving unit 10 derives a fade-out coefficient in order to adjust the volume based on the noise component determination signal output from the noise determination unit 9. The fade-out coefficient deriving unit 10 may derive the fade-out coefficient by calculating the noise component based on the determination signal of the noise component. Alternatively, a fade-out coefficient corresponding to the noise component determination signal is set in advance and this is set in a table, and the noise component determination signal output from the noise determination means 9 is applied to the table to derive the fade-out coefficient. May be.

  The sound volume adjusting means 11 is inputted with the digital input signal of the voice band and the fade-out coefficient signal derived by the fade-out coefficient deriving means 10. The volume adjusting means 11 adjusts the level of the digital input signal in the audio band based on the fade-out coefficient, and outputs an audio signal whose volume is adjusted according to the magnitude of the fade-out coefficient. FIG. 4 is an example of a fade-out characteristic with respect to the reception level of the infrared light receiving unit in the infrared wireless receiver. A dotted curve indicates the amount of noise, a broken curve indicates the fade-out amount, and a solid curve indicates the output volume. The horizontal axis represents the reception level (dBμV), and the vertical axis represents the noise amount, the fade-out amount, and the output volume in percentage (%). As apparent from FIG. 4, if the reception level is a certain level or more, the amount of noise is negligible and there is no need to fade out, so the output volume is maintained at 100%. When the reception level is 30.0 (dBμV) or less, the ratio of the noise component increases so much that it cannot be ignored. Therefore, the fade-out amount is increased from this area, and the output volume is lowered as the fade-out amount increases. It has become.

  The signal whose volume is adjusted by the volume adjusting means 11 and output is converted into an analog audio signal by the D / A converter, and the sound is emitted from the speaker via an amplifier and other appropriate circuits.

  The volume control means 11 can be configured by an attenuator, for example. The fade-out characteristic shown in FIG. 4, that is, the relationship between the fade-out amount and the noise amount is merely an example, and the fade-out characteristic may be appropriately adjusted depending on the purpose of use, communication method, required communication quality, and the like. For example, the fade-out characteristic can be adjusted by adjusting the attenuation characteristic of the attenuator constituting the volume control means 11.

  The volume control means 11 may be configured to continuously adjust the volume, that is, the output level of the audio signal, according to the magnitude of the fade-out coefficient, or may be adjusted in stages.

  Each block shown in FIG. 1 may physically exist for each of these blocks, but all functions performed by each block are given to a digital signal processor (DSP). Good.

  According to the first embodiment described above, even a digital audio processing device can realize so-called soft mute without a sense of incongruity by changing the fade-out amount according to the reception state.

  Next, a second embodiment shown in FIG. 2 will be described. In the second embodiment, the noise reduction means 4, the second fast Fourier transform means 5, the second absolute value calculation means 6, and the sound restoration means 7 are added to the configuration of the first embodiment shown in FIG. is there. Since the signal transmission path is slightly different from that of the first embodiment, the different components will be described mainly.

  In FIG. 2, the digital input signal of the audio band signal is a fast Fourier transform means 1, an absolute value calculation means 2, an autocorrelation function calculation means 3, a difference calculation means 8, a noise determination means 9, a fade-out coefficient derivation means 10, a volume control. It is the same as the first embodiment in that a predetermined process is performed by each means by passing through the means 11 in order. The fast Fourier transform unit 1 is referred to as “first fast Fourier transform unit 1” in order to distinguish it from the second fast Fourier transform unit 5. Similarly, the absolute value calculation means 2 is referred to as “first absolute value calculation means 2”.

  The autocorrelation function signal output from the autocorrelation function calculation unit 3 is the same as that of the first embodiment in that the autocorrelation function signal is input to the difference calculation unit 8. The point of being configured to input is different from the first embodiment. The noise reduction means 4 is composed of, for example, a low-pass filter and removes high-frequency components from the autocorrelation function signal. This reduces non-periodic noise components such as white noise from the frequency domain signal.

  The output signal of the noise reduction means 4 is input to the second fast Fourier transform means 5, and the output signal of the noise reduction means 4 is fast Fourier transformed to be returned to the frequency domain signal. The signal in the frequency domain is represented by a complex number. In order to calculate the gain from the signal represented by the complex number, the second absolute value calculating unit 6 calculates an absolute value.

  The absolute value signal of the frequency domain signal calculated by the second absolute value calculation means 6, that is, the gain signal of the frequency domain signal is input to the sound restoration means 7. The audio restoration means 7 receives a phase signal obtained by converting the digital audio band input signal from the time domain signal to the frequency domain signal in the first fast Fourier transform means 1. The audio restoration means 7 restores the digital audio band signal by combining the gain signal and the frequency domain signal. As described above, the components from the noise reduction unit 4 to the second fast Fourier transform unit 5, the second absolute value calculation unit 6, and the audio restoration unit 7 are connected to the digital audio band from the output of the autocorrelation function calculation unit 3. A circuit for restoring and generating the signal is configured. The digital audio band signal restored in this way is input to the volume control means 11.

  The circuit configuration from the difference calculating unit 8 to the noise determining unit 9, the fade-out coefficient deriving unit 10, and the volume adjusting unit 11 is the same as the circuit configuration in the first embodiment. The difference calculation means 8 performs high-pass filter processing to obtain the amplitude difference, that is, the amount of change of the correlation waveform included in the autocorrelation function obtained by the autocorrelation function calculation means 3, and a signal having a large amount of non-periodic noise signal components It is determined whether or not. The fade-out coefficient deriving unit 10 derives a fade-out coefficient in order to adjust the volume based on the noise component determination signal output from the noise determination unit 9. The volume control unit 11 adjusts the level of the digital input signal in the audio band based on the fade-out coefficient, and adjusts the fade-out amount according to the magnitude of the fade-out coefficient. The volume adjusting means 11 adjusts the level of the digital input signal in the audio band based on the fade-out coefficient, and outputs an audio signal whose volume is adjusted according to the magnitude of the fade-out coefficient.

  Similar to the first embodiment, the signal whose volume is adjusted by the volume adjusting means 11 and output is converted into an analog audio signal by the D / A converter, and the sound is output from the speaker via an amplifier and other appropriate circuits. Released. Moreover, the volume control means 11 can be comprised with an attenuator (attenuator), for example. Further, all the functional blocks shown in FIG. 2 may be constituted by a digital signal processor (DSP). The volume adjustment by the volume adjustment unit 11 may be performed continuously or stepwise depending on the magnitude of the fade-out coefficient.

  According to the second embodiment, in addition to the same effects as those of the first embodiment, the noise component included in the digital input signal in the audio band is reduced, so that an audio output with less noise can be obtained.

  The present invention is suitable as an audio processing apparatus in various apparatuses or systems for transmitting and receiving audio band signals wirelessly and reproducing audio, for example, an audio conference system using radio waves or light as a transmission medium, broadcasting or a communication system.

DESCRIPTION OF SYMBOLS 1 Fast Fourier transform means 2 Absolute value calculation means 3 Autocorrelation function calculation means 4 Noise reduction means 5 Second fast Fourier transform means 6 Second absolute value calculation means 7 Speech restoration means 8 Difference calculation means 9 Noise determination means 10 Fade out Coefficient derivation means 11 Volume control means

Claims (6)

A fast Fourier transform means for performing a Fourier transform of a digital input signal in a voice band at a high speed;
Absolute value calculating means for generating an absolute value signal by calculating an absolute value of the Fourier transformed signal;
Autocorrelation function calculating means for calculating an autocorrelation function of the calculated absolute value signal;
Difference calculating means for calculating the amount of change in output of the autocorrelation function calculating means;
Noise determining means for calculating a noise component according to the magnitude of the difference;
A fade-out coefficient deriving unit that calculates a fade-out amount based on the calculation result by the noise determination unit;
Volume adjusting means for adjusting and outputting the level of the digital input signal based on the calculation result by the fade-out coefficient deriving means;
A speech processing apparatus comprising: First fast Fourier transform means for fast Fourier transform of a digital input signal in a voice band;
First absolute value calculating means for generating an absolute value signal by calculating an absolute value of the Fourier transformed signal;
Autocorrelation function calculating means for calculating an autocorrelation function of the calculated absolute value signal;
Difference calculating means for calculating the amount of change in output of the autocorrelation function calculating means;
Noise determining means for calculating a noise component according to the magnitude of the difference;
A fade-out coefficient deriving unit that calculates a fade-out amount based on the calculation result by the noise determination unit;
Second fast Fourier transform means for fast Fourier transforming the output of the autocorrelation function calculating means;
Second absolute value calculating means for generating an absolute value signal by calculating an absolute value of the signal subjected to Fourier transform by the second fast Fourier transform means;
Speech restoration means for restoring speech from a phase signal obtained by fast Fourier transforming the digital input signal by the first fast Fourier transform means and an absolute value signal output from the second absolute value calculation means;
Volume adjusting means for adjusting and outputting the level of the audio signal restored by the audio restoring means based on the calculation result by the fade-out coefficient deriving means;
A speech processing apparatus comprising:   The speech processing apparatus according to claim 2, wherein a noise reduction means is provided between the autocorrelation function calculation means and the second fast Fourier transform means.   4. The speech processing apparatus according to claim 3, wherein the noise reduction means is a filter that removes a high frequency component of the function calculated by the autocorrelation function calculation means.   The sound processing apparatus according to claim 1, wherein the volume adjusting means continuously adjusts the level of the sound signal. The sound processing apparatus according to claim 1, wherein the volume adjusting means adjusts the level of the sound signal stepwise.

Images