JP3947021B2 - Call voice processing device - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a technique for improving the clarity of call voice in voice communication.
[0002]
[Prior art]
There is a form called a hands-free call as a call form of a user using a telephone which is a typical example of voice communication.
The hands-free call is a call using a telephone using a microphone and a speaker outside the telephone for input of transmitted voice and output of received voice. Here, a device that adds the above-described hands-free call function to a mobile phone is called a hands-free call device, etc., and in a car, a user who is driving can use the mobile phone to perform a hands-free call in the car space. It is used for making it possible to perform via a speaker that radiates received voice and a microphone that picks up voice in the interior of the vehicle.
[0003]
[Problems to be solved by the invention]
For example, in a hands-free call in an automobile as described above, the reception sound output from the speaker is greatly circulated into the microphone (echo), and the traveling sound or music output from the in-vehicle audio device, etc. Since the received voice output from the speaker is difficult to hear due to ambient noise, the call voice is likely to be obscured compared to a call using a normal telephone handset.
[0004]
Therefore, an object of the present invention is to efficiently suppress obscuration of call voice in voice communication even in an environment where both echo and ambient noise are relatively large.
[0005]
[Means for Solving the Problems]
In order to achieve the above object, the present invention provides a call voice processing apparatus for processing a call voice of a voice communication terminal, a loudness compensation means for compensating a loudness of received voice received by the voice communication terminal, and a loudness compensation means. Echo cancellation for canceling echo of received voice included in voice input from microphone, speaker for outputting received voice subjected to loudness compensation, voice input from microphone, and transmitting voice transmitted by said voice communication terminal Means for calculating the estimated echo of the received voice included in the voice input from the microphone as the estimated echo voice, and the estimated echo calculation means from the voice input from the microphone. Subtract the estimated echo sound calculated by Constructed by including a calculation unit, in said loudness compensation means, estimated echo sound the estimated echo calculation means has calculated So that the clarity of the received voice recognized by the user is maintained by regarding the transmitted voice as the ambient noise of the user directly. The gain of the received voice is adjusted.
[0006]
According to such a call voice processing device, it is possible to suppress (echo cancellation) the mixing of echoes from the received voice into the transmitted voice. In addition, the correlation between the estimated echo sound calculated during the echo cancellation process and the received voice reaching the user and the correlation between the transmitted voice after echo cancellation and the user's ambient noise can be expected to be strong. Estimated echo sound So that the clarity of the received voice recognized by the user is maintained by regarding the transmitted voice as the ambient noise of the user directly. By appropriately adjusting the gain of the received voice, it is derived based on the loudness theory that the obscuration due to the ambient noise of the received voice for the user can be suppressed. Therefore, according to the call voice processing apparatus, the call voice is obscured by both echo and ambient noise with a simple and efficient configuration in which part of the configuration for echo cancellation and the configuration for loudness compensation are shared. Deterrence can be realized.
[0008]
Further, in the loudness compensation means, for each predetermined frequency band, the received voice according to the estimated echo voice calculated by the estimated echo calculation means in the frequency band and the transmitted voice in the frequency band. If the gain adjustment is performed, it is possible to realize ease of listening that does not depend on the frequency of the received voice of the user.
[0009]
Here, such a call voice processing apparatus is excellent in application to a call voice processing apparatus having a mobile telephone interface that accommodates a mobile telephone as the voice communication terminal, that is, a hands-free apparatus.
[0010]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, an embodiment of a call voice processing device according to the present invention will be described taking application to an in-vehicle hands-free call device as an example.
FIG. 1 shows the configuration of a hands-free call device according to the present embodiment.
As shown in the figure, the hands-free call device includes a voice processing device 1, a speaker 2, a microphone 3, a telephone interface device 4, and a volume switch 5. A mobile phone 6 is connected to the telephone interface device 4.
In such a configuration, the telephone interface device 4 is detachably connected to the mobile phone 6, outputs a voice signal of the received voice input from the mobile phone 6 to the voice processing device 1, and transmits it from the voice processing device 1. The speech signal is output to the mobile phone 6.
In addition, the telephone interface device 4 includes an input device such as a switch, relays an on-hook / off-hook command to the mobile phone 6 according to a user operation of the input device, and causes the mobile phone 6 to perform an on-hook / off-hook operation. Also do.
[0011]
Next, the audio processing device 1 is roughly composed of an echo canceller unit 11 and an audio output processing unit 12. The echo canceller unit 11 includes an estimated echo calculation unit 111 and an adder 112, and the estimated echo calculation unit 111 includes an adaptive filter 1111 and a coefficient update unit 1112. On the other hand, the audio output processing unit 12 includes a loudness compensation control unit 121 and an audio adjustment unit 122, and the audio adjustment unit 122 includes a volume adjustment unit 1221, a frequency gain adjustment unit 1222, and an output amplifier 1223.
[0012]
In such a configuration, the volume adjustment unit 1221 adjusts and outputs the received voice input from the telephone interface device 4 to a gain according to the setting by the user's volume switch 5. Then, the frequency gain adjustment unit 1222 adjusts the gain for each frequency band of the reception signal output from the audio adjustment unit 122 according to the setting from the loudness compensation control unit 121, and outputs it to the output amplifier 1223. The output amplifier 1223 drives the speaker 2 with a signal obtained by amplifying the received reception signal, and outputs sound to the vehicle interior space.
[0013]
On the other hand, the microphone 3 picks up the voice of the interior space and sends it to the adder 112 as a transmission signal. The adder 112 subtracts the estimated echo signal calculated by the estimated echo calculation unit 111 from the transmission signal, and then outputs it to the telephone interface device 4 as a transmission signal after echo cancellation.
[0014]
Hereinafter, the echo canceling operation for canceling the echo of the reception signal included in the transmission signal picked up by the microphone 3 and the gain of each frequency band of the reception signal output from the speaker 2 in the speech processing apparatus 1 as described above are described as in-vehicle noise. The loudness compensation operation for compensating the user so that the user can easily hear is explained.
[0015]
First, the echo cancellation operation will be described.
Now, s (k) is an audio signal input from the audio interface, g (k) is an impulse response of the transmission system of the audio adjustment unit 122, and h (k) is an impulse response of the transmission system from the speaker input to the microphone output. Then, the echo component of the received signal included in the transmitted signal output from the microphone 3 is expressed as d (k) = {h (k) * g (k) * s (k)}. Here, * represents a convolution operation. Further, the microphone output component is expressed as d (k) + n (k), where n (k) is the surrounding voice other than the echo of the reception signal output from the microphone 3. Therefore, the estimated echo calculation unit 111 estimates the impulse response g (k) * h (k) of the transmission system from the sound adjustment unit input to the microphone output, sets it as the impulse response of the adaptive filter 1111, and outputs the signal s (k ) To the adaptive filter 1111, the echo component of the received signal mixed in the microphone output component can be generated as the estimated echo signal y (k). Then, when the adder 112 subtracts the estimated echo signal y (k) from the transmission signal d (k) + n (k) output from the microphone 3, the transmission signal d (k) output from the microphone 3 is obtained. ) + N (k), it is possible to obtain a transmission signal e (k) in which the echo component d (k) of the reception signal included in the received signal is canceled.
[0016]
Here, the coefficient update unit 1112 calculates and updates the impulse response of the adaptive filter 1111. The coefficient updating unit 1112 uses a filter identification coefficient of the adaptive filter 1111 so that the power of the adder output e (k) is minimized by a learning identification method such as an LMS (Least Mean Square) algorithm or an NLMS (Normalized Least Mean Square) algorithm. And the process of setting the adaptive filter 1111 is repeated. However, when the level of the received signal s (k) is small or the fluctuation of the ambient sound n (k) is severe, the impulse response g (k) * h ( k) cannot be estimated satisfactorily, so when the level of the received signal s (k) is lower than a predetermined level or when the level of the surrounding voice n (k) fluctuates during a user's speech, The calculation and update of the filter coefficient of the adaptive filter 1111 may be stopped. Here, the fluctuation of the ambient voice n (k) appears as the fluctuation of the transmission signal e (k). Alternatively, the filter coefficient can be calculated and updated with a period as long as the fluctuation of the ambient sound n (k) is canceled.
[0017]
As an echo cancellation technique using such an adaptive filter 1111, Japanese Patent Laid-Open No. 2000-341178, Japanese Patent Laid-Open No. 2-288428, Japanese Patent Laid-Open No. 7-086991, Japanese Patent Laid-Open No. 10-257583, etc. The described techniques can be used. The adaptive filter 1111 is, for example, an FIR (Finite Impulse Response) filter. In this case, the impulse response of the adaptive filter 1111 is performed by setting the tap coefficient Xj of the FIR filter.
[0018]
The operation of the echo cancellation operation has been described above.
Next, the loudness compensation operation will be described.
First, in the loudness compensation operation, the principle of how to make it easy for the user to hear the received voice will be described.
The unit of “the volume of sound perceived by humans (loudness)” is “sone”, and the volume of a pure sound of 1 kHz and 40 dB is 1 “sone”. Because it is based on human perception, 2 sounds sounds twice as large as 1 sound. Loudness changes not only with sound intensity but also with frequency band. FIG. 2 is an equal loudness level curve obtained by connecting the sound pressure levels of pure tones having the same loudness as that of a pure tone having a sound pressure level of 1 kHz without external noise. That is, the equal loudness level curve is a plot of the level of another frequency at which a person sounds as large as a 1 kHz sine wave. The equal loudness level curve indicates that as the level decreases, the sound in the low frequency range and the high frequency range may be heard lower or lower than the sound in the intermediate frequency range unless the level is raised.
[0019]
Next, FIG. 3 shows the correspondence between the physical sound pressure level and the loudness felt when a person is listening to the sound, and is called a loudness curve. In the loudness curve, the horizontal axis is the physical sound pressure level (unit: Sound Pressure Level; SPL (dB)), and the vertical axis is the loudness (unit: sone), which is the numerical value of the sound perceived by humans. . In FIG. 3, (a) is a loudness curve in a quiet environment, and (b) is a loudness curve under noise. Note that (b) is a curve in a noise in which the minimum audible value of a person increases by about 35 dB, and this curve also changes variously as the noise changes.
[0020]
Here, the loudness curve indicates that if the numerical value of the loudness on the vertical axis is the same, a person feels that the sound has the same loudness. Therefore, the sound that a person feels at a magnitude of 0.1 zone may be a physical sound pressure level of 12 dB SPL in the quiet environment of (a), but a physical sound pressure level of 37 dB SPL in the noise of (b). is required. In other words, if a 12 dB SPL sound is output from the speaker 2 in a quiet environment, the sound of the same magnitude may be felt if the 37 dB SPL sound is not output from the speaker 2 under the noise of (b). Can not. In other words, in order to hear a sound that feels as small as 0.1 zone under noise, a gain of 25 dB must be added compared to listening in a quiet environment. In addition, the sound that a person feels at a 1-sone level is a physical sound pressure level of 42 dB SPL in the quiet environment of (a), but a physical sound pressure level of 49 dB SPL is necessary under the noise of (b). So we have to add a gain of 7dB.
[0021]
Here, when the sound that the user wants to hear is the output sound of the speaker, in order to make the sound output by the speaker feel as a constant loudness regardless of the noise level, It is necessary to change the gain depending not only on the level but also on the level of the sound output from the speaker. Here, FIG. 4 is a diagram showing how much gain needs to be added to the sound pressure level under silence in order to feel a sound of the same magnitude as under silence under noise. In the figure, the horizontal axis represents the sound pressure level of the sound output under silence, and the vertical axis represents the gain value that needs to be added in order to feel a sound of the same magnitude as under silence. For example, a sound that is output at a sound pressure level of 20 dB under silence is added with a gain of about 19 dB under noise, so that a human feels that the sound has the same magnitude as that under silence.
[0022]
As described above, the gain that needs to be given to the speaker output sound differs depending on the ambient noise level and the speaker output sound level in order to realize the same ease of hearing for the user. In addition, the ambient noise has a different level for each frequency band, and the ease of listening to the sound of the user differs for each frequency band as shown in the equal loudness level curve in FIG. In order to realize the same ease of listening, the gain that needs to be given to the speaker output sound needs to be different for each frequency band.
[0023]
Therefore, in this embodiment, a gain adjustment amount that realizes ease of hearing regardless of the ambient noise level and the frequency band is determined for each combination of the received voice level and the ambient noise level for each frequency band, and the loudness compensation control is performed. The unit 121 estimates the received voice level and the ambient noise level output to the user for each frequency band, and selects a gain adjustment amount set in advance for a set of the estimated received voice level and the ambient noise level, According to the gain adjustment amount selected for each frequency band, the frequency gain adjustment unit 1222 adjusts the gain of the received voice for each frequency band. In this embodiment, the level of the estimated echo signal y (x) estimated by the estimated echo calculation unit 111 described above is used as the received voice level, and the transmission signal e (x) after echo cancellation is used as the ambient noise level. By using it, the received voice level and ambient noise level output to the user are estimated. Here, when the estimation of the echo by the estimated echo signal calculation unit is appropriately performed, the user can be expected to be relatively close to the microphone, so the estimated echo signal y (x) is the sound of the received voice that reaches the user. It can be considered that the transmission signal e (x) after echo cancellation is substantially proportional to the pressure level, and is proportional to the sound pressure level of noise around the user at least when the user is not speaking.
[0024]
Details of such a loudness compensation operation will be described below.
FIG. 6 shows a configuration example of the loudness compensation control unit 121.
As shown in the figure, the loudness compensation control unit 121 includes an ambient noise frequency band level averaging unit 1211, a noise level correction unit 1212, a frequency band gain table selection unit 1213, a received voice frequency band level averaging unit 1214, and a gain table memory 1215. It is configured.
[0025]
In the gain table memory 1215, a gain table that preliminarily describes the relationship between the received voice level and the gain to be added and is provided for each combination of various noise levels and frequency bands in advance, for example, defining the relationship as shown in FIG. Is recorded.
[0026]
The ambient noise frequency band level averaging unit 1211 performs an FFT (Fast Fourier Transform) operation for each predetermined time block on the transmission signal e (x) after echo cancellation output from the adder 112, The average sound pressure level in the time block is calculated for each frequency band. Here, for example, in consideration of the characteristic that human hearing can recognize the difference in noise level every 1/3 octave, the average sound within the time block for each frequency band of 1/3 octave. Calculate the pressure level.
[0027]
The noise level correction unit 1212 adjusts the sound pressure level output for each frequency band from the ambient noise frequency band level averaging unit 1211 using Zwicker's loudness calculation method (ISO 532B) or Stevens' loudness calculation method (ISO 532A). To do.
Specifically, the adjustment is performed as follows. That is, when there is noise of a certain frequency component, this noise etc. not only affects the difficulty of listening to the received voice of the same frequency component, but also listens to the guide voice signal of the frequency component adjacent to the high frequency side. It also affects the difficulty. In view of this, the noise level correcting unit 1212 adjusts the sound pressure level of each frequency component according to the magnitude of the sound pressure level of the frequency component such as noise adjacent to the low frequency side. That is, when the sound pressure level of the adjacent low frequency component is large, the sound pressure level of the frequency component adjacent to the high frequency side is corrected to be higher. By making such adjustments, when selecting a gain table for each frequency band, it is only necessary to focus on the sound pressure level such as noise in each corresponding frequency band, and the frequency band adjacent to the low frequency side. Therefore, it is not necessary to perform complicated processing such as taking into account noise and the like.
[0028]
The received voice frequency band level averaging unit 1214 performs a well-known FFT operation on the estimated echo signal output from the estimated echo calculation unit 111 for each predetermined time block, and the received noise frequency band level averaging unit 1211 The average sound pressure level in the time block is calculated for each frequency band.
[0029]
Then, the frequency band gain table selection unit 1213 selects, for each frequency band, a gain table corresponding to the frequency band and the sound pressure level of the adjusted frequency band output from the noise level correction unit 1212. . Then, for each frequency band, using the selected gain table, a gain value corresponding to the sound pressure level of the frequency band output from the received voice frequency band level averaging unit 1214 is calculated, and the frequency gain adjustment unit 1222 receives the gain value. Sent.
[0030]
The loudness compensation control unit 121 has been described above.
It should be noted that the ambient noise frequency band level average unit 1211 and the received voice frequency band level average unit 1214 of the present loudness compensation control unit are divided into a set of frequency filters that divide the signal for each frequency band, instead of the FET calculation. In addition, it may be configured with a time averaging unit for obtaining a time average of sound pressure levels of signals in each frequency band. Alternatively, the ambient noise frequency band level average unit 1211 and the received voice frequency band level average unit 1214 of the present loudness compensation control unit increase the frequency resolution by increasing the time window length in the lower frequency band instead of the FET calculation. It is also possible to use a window function with a reduced time resolution and perform a wavelet transform using a window function with a reduced frequency resolution by shortening the time window length and increasing the time resolution for higher frequency bands.
[0031]
Next, the frequency gain adjustment unit 1222 will be described.
FIG. 6 shows a configuration example of the frequency gain adjustment unit 1222.
In the illustrated example, the frequency gain adjustment unit 1222 includes a filter bank 12221, a variable gain unit 12222, and an adder 12223.
The filter bank 12221 is a band pass filter group having a predetermined frequency bandwidth, and the received voice signal is divided into frequency bands by these band pass filter groups. The variable gain unit 12222 gives the gain for each frequency band calculated by the loudness compensation control unit 121 to the received voice signal divided for each frequency band output from the filter bank 12221 to perform gain adjustment. The adder 12223 adds and outputs the received voice signal whose gain is adjusted for each frequency band.
[0032]
The frequency gain adjustment unit 1222 has been described above.
The frequency gain adjustment unit 1222 uses the FIR filter and the gain of each frequency band calculated by the loudness compensation control unit 121 as the gain of the center frequency of each frequency band, and uses a spline function between the gain values. The spline function interpolator that calculates smooth gain characteristics in the frequency domain by interpolating using, and the gain characteristics calculated by the spline function interpolator are converted from the frequency domain to the time domain using IFFT (InverseFast Fourier Transform) The IFFT calculation unit for setting the value of the tap coefficient of the FIR filter can also be configured. In this case, the FIR filter performs filtering processing on the time axis for the received voice signal, and performs gain adjustment similar to the frequency gain adjustment unit 1222 illustrated in FIG.
[0033]
Further, the frequency gain adjustment unit 1222 uses the gain of each frequency band calculated by the loudness compensation control unit 121 as the gain of the center frequency of each frequency band, and interpolates between each gain value using a spline function. Spline function interpolation unit that calculates smooth gain characteristics in the frequency domain, FFT operation unit that performs FFT operation on the received voice signal and converts it from the time domain to the frequency domain, and the frequency domain that is output from the FFT operation unit The frequency band filtering unit that filters the guidance voice signal at the frequency band filtering unit that performs smooth gain characteristics output from the spline function interpolation unit and the IFFT operation on the received voice signal in the frequency domain that is output from the frequency band filtering unit IFFT calculator that converts frequency domain to time domain It is also possible to more configuration. In the IFFT calculation process, a well-known superposition addition method (overlap-addmethod) or superposition hold method (overlap-save method) may be used in order to realize linear filtering. With this configuration, the amount of calculation can be relatively reduced even when the number of filter taps is large.
[0034]
In any case, it is preferable that the frequency gain adjusting unit 1222 gradually updates the gain characteristics because the output waveform becomes discontinuous when the gain changes abruptly.
Heretofore, an embodiment of the audio processing device 1 according to the present invention has been described.
In the above, application to an in-vehicle hands-free communication device has been described as an example, but the configuration for performing echo cancellation and loudness compensation as described above is a mobile phone, a fixed phone or any other voice communication device, or The present invention can be applied to a device that processes call voice of any voice communication device.
[0035]
Also, in the above, the voice band is divided into a plurality of frequency bands, and the gain of the received voice is adjusted for each frequency band. Also good. In this case, the frequency gain adjusting unit 1222 can be omitted, and it is sufficient to adjust the gain of the received voice by adjusting the volume of the volume adjusting unit 1221.
[0036]
As described above, according to the present embodiment, it is possible to suppress obscuration of call speech due to both echo and ambient noise in voice communication. In addition, the estimated echo signal calculated in the echo cancellation process is regarded as a received voice to be clarified in loudness compensation, and the transmitted signal after echo cancellation is regarded as ambient noise in loudness compensation, and loudness compensation is performed. As a result, a simple and efficient configuration that shares part of the configuration for echo cancellation and the configuration for loudness compensation prevents suppression of obscuration of call speech due to both echo and ambient noise. Realized.
[0037]
However, in the loudness compensation, the estimated echo signal calculated in the process of echo cancellation is clarified in the loudness compensation by performing a predetermined correction to reduce the deviation between the estimated echo signal and the received signal reaching the actual user. A signal that has been subjected to a predetermined correction to reduce the difference between the echo signal after echo cancellation and the actual ambient noise of the user. May be regarded as ambient noise in the loudness compensation, and the loudness compensation may be performed.
[0038]
【The invention's effect】
As described above, according to the present invention, even in an environment where both echo and ambient noise are relatively large, it is possible to efficiently suppress obscuration of call voice in voice communication.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of a hands-free device according to an embodiment of the present invention.
FIG. 2 is a diagram showing an equal loudness level curve.
FIG. 3 is a diagram showing a loudness curve in a quiet environment and a noisy environment.
FIG. 4 is a diagram illustrating a gain for obtaining the same loudness in a quiet environment and a noisy environment.
FIG. 5 is a block diagram illustrating a configuration example of a loudness compensation control unit according to the embodiment of the present invention.
FIG. 6 is a block diagram illustrating a configuration example of a frequency gain adjustment unit according to the embodiment of the present invention.
[Explanation of symbols]
1: audio processing device, 2: speaker, 3: microphone, 4: telephone interface device, 5: volume switch, 6: mobile phone, 11: echo canceller unit, 12: audio output processing unit, 111: estimated echo calculation unit, 112: Adder, 121: Loudness compensation control unit, 122: Audio adjustment unit, 1111: Adaptive filter, 1112: Coefficient update unit, 1211: Ambient noise frequency band level averaging unit, 1212: Noise level correction unit, 1213: Frequency band Gain table selection unit, 1214: Received voice frequency band level averaging unit, 1215: Gain table memory, 1221: Volume adjustment unit, 1222: Frequency gain adjustment unit, 1223: Output amplifier, 12221: Filter bank, 12222: Variable gain unit, 12223: Adder.

Claims (4)

A call voice processing device for processing call voice of a voice communication terminal,
Loudness compensation means for compensating the loudness of the received voice received by the voice communication terminal; and a speaker for outputting the received voice subjected to the loudness compensation by the loudness compensation means;
A microphone for voice input,
Canceling echo of the received voice included in the voice input from the microphone, and having an echo canceling means for transmitting the voice transmitted by the voice communication terminal,
The echo canceling means is an estimated echo calculating means for calculating an estimated value of an echo of a received voice included in a voice input from a microphone as an estimated echo voice;
Subtracting the estimated echo sound calculated by the estimated echo calculation means from the sound input from the microphone, and the subtracting means as the transmission voice,
The loudness compensation means maintains the clarity of the received voice recognized by the user by using the estimated echo sound calculated by the estimated echo calculating means as the received voice that reaches the user and using the transmitted voice as the ambient noise of the user. Further, the call voice processing apparatus is characterized in that the gain of the received voice is adjusted. The call voice processing device according to claim 1,
The loudness compensation means adjusts the gain of the received voice for each predetermined frequency band according to the estimated echo sound calculated by the estimated echo calculation means in the frequency band and the transmitted voice in the frequency band. A call voice processing apparatus characterized by: The call voice processing device according to claim 1 or 2 ,
A call voice processing apparatus comprising a mobile telephone interface for accommodating a mobile telephone as the voice communication terminal. A call voice processing method for processing call voice in voice communication,
The estimated value of the echo of the received voice included in the voice input from the microphone is calculated as the estimated echo voice, and the estimated echo voice calculated from the voice input from the microphone is subtracted to obtain the transmitted voice to be transmitted in voice communication. Steps,
The estimated echo sound as the received voice to reach the user, the transmitting voice as ambient noise of a user, call voice processing method characterized by a step of outputting from the speaker by performing the gain adjustment of the received voice.

Images