Voice communication device and program

Abstract

To provide a voice communication device that has a speaker and a plurality of microphones and performs voice communication from a near end to a far end, which can impart directionality to a voice signal input to the plurality of microphones and can suppress echoes caused by voice output from the speaker.SOLUTION: A beamformer portion, which gives directivity to an audio signal, estimates the characteristics between a plurality of microphones using an adaptive filter, expands the estimated filter coefficients in the frequency domain, calculates a phase difference between the microphones, and gives directivity. The beamformer portion also performs audio processing on a path from the near end to the far end so as to suppress noise in a predetermined width of the frequency domain, suppress echoes generated by the audio signal output from the speaker, and further suppress howling.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a voice communication device and a program. [Background technology]

[0002] In order to enable conversations between different spaces, there are voice communication devices equipped with a speaker that amplifies the voice of the other speaker and a microphone that inputs the voice. In particular, in devices that support hands-free calling, the voice amplified from the speaker is input to the microphone, and the voice of the other speaker is heard with a delay due to echo (reverberation), which interferes with the conversation. Therefore, an echo canceller is used to cancel the echo that occurs between the speaker and the microphone. Also, when noise around the microphone is unpleasant, it is common to perform noise suppression.

[0003] An echo canceller achieves echo attenuation by estimating a linear filter from the speaker and microphone signals using an adaptive filter and subtracting it from the signal from the microphone. However, if the microphone signal contains many components that are not derived from the speaker, this hinders echo attenuation. For this reason, a method has been proposed in which the method of inserting the amount of insertion loss is changed depending on the level of background noise and the difficulty of echo attenuation (see, for example, Patent Document 1).

[0004] Also, as a voice communication device for conversation in the same space, a voice communication device for hands-free conversation between a driver and a person in the back seat of a car has been proposed (Patent Document 2). Furthermore, a method for performing sound source localization and sound source separation, called a beamformer, is known (Non-Patent Document 1). [Prior art documents] [Patent documents]

[0005] [Patent Document 1] JP 2007-324675 A [Patent Document 2] JP 2015-713210 A [Non-Patent Document 1] Futoshi Asano, Acoustic Technology Series 1, “Array Signal Processing of Sound - Localization, Tracking and Separation of Sound Sources”, edited by the Acoustical Society of Japan, Corona Publishing, 2011 Summary of the Invention [Problem to be solved by the invention]

[0006] There are cases where two voice communication devices are placed in one space to enable hands-free communication. In places where such a request is made, there are phenomena such as background noise interfering with the conversation or difficulty in hearing live voices. In addition, when multiple voice communication devices that can be used for conversation in different spaces are connected in one space, if noise occurs in one space, the noise is input to the microphones of both devices, causing problems such as not being able to achieve the echo attenuation expected of the echo canceller or causing howling.

[0007] In addition, the voice communication device proposed in Patent Document 2 for hands-free calling inside a car uses a camera device to determine whether the speaker is speaking into a microphone, and then identifies the speaker. This requires information processing to determine the speaker from the camera device or captured images, which complicates the processing and increases the cost of the device.

[0008] The present invention has been made in view of such problems, and aims to provide a voice communication device that can maintain a call state between speakers who are located at different positions in the same space, even when the other person's voice is difficult to hear due to noise or distance. Also, the present invention aims to provide a voice communication device and a program that can realize beamforming that gives directionality to pass the speaker's voice without using a camera device or the like. [Means for solving the problem]

[0009] One aspect of the present invention is a voice communication device having a plurality of microphones, one of which is a reference microphone, and a beamformer unit that imparts directivity to voice input to the plurality of microphones, passing voice arriving from a desired direction and suppressing voice arriving from other directions, wherein the beamformer unit is characterized by comprising: a delay analysis unit which is an adaptive filter that estimates an inter-microphone characteristic filter coefficient using voice signals input from the plurality of microphones; a directivity analysis unit that converts the estimated inter-microphone characteristic filter coefficient into the frequency domain, detects a phase difference between the voice signals input from the plurality of microphones, and determines the directional direction of the passband or stopband of the voice input to the plurality of microphones; a directional filter unit that passes voice in the pass direction determined by the directivity analysis unit, generates a filter coefficient that imparts directivity that blocks voice in the block direction, converts it into the time domain, and weights voice signals from microphones other than the reference microphone with the converted filter coefficient; and a subtraction unit that subtracts the voice signals from the other microphones that have been weighted by the directional filter unit from the voice signals input from the reference microphone to pass the voice signals in the pass direction.

[0010] Another aspect of the present invention is a voice communication device which outputs an audio signal input to a sending input terminal from a sending output terminal and outputs an audio signal input to a receiving input terminal from a speaker of the receiving output terminal, and which is provided between the sending input terminal and the sending output terminal and which suppresses echoes generated by the audio signal output from the receiving output terminal, wherein the echo canceller comprises a noise suppression unit which performs frequency conversion on the audio signal input to the sending input terminal to detect noise in a predetermined frequency range and suppresses noise components for each predetermined frequency range, an echo suppression unit which has an adaptive filter which estimates a pseudo echo signal due to the received audio signal output from the speaker of the receiving output terminal and suppresses echoes from the audio signal in which the noise components have been suppressed by the noise suppression unit, and a howling suppression unit which performs howling suppression processing on the echo-suppressed audio signal.

[0011] Another aspect of the present invention is a voice communication device having an echo canceller that outputs an audio signal input to the above-mentioned sending input terminal from a sending output terminal and outputs an audio signal input to the receiving input terminal from a speaker of the receiving output terminal, and is provided between the sending input terminal and the sending output terminal and suppresses echoes caused by the audio signal output from the receiving output terminal, the voice communication device further comprising a plurality of microphones, one of the plurality of microphones being a reference microphone, and a beamformer unit that imparts directivity to audio input to the plurality of microphones so as to pass audio arriving from a desired direction and suppress audio arriving from other directions, the beamformer unit having an adaptive filter that estimates an inter-microphone characteristic filter coefficient indicating the characteristics of the audio input to the other microphones relative to the audio input to the reference microphone, frequency-converts the estimated inter-microphone characteristic filter coefficient, detects a phase difference between the audio signals from the plurality of microphones from the converted inter-microphone characteristic filter coefficient, and imparts directivity to the audio signals from the plurality of microphones based on the phase difference. Effect of the Invention

[0012] According to the present invention, since it is possible to impart directivity to an audio signal by using an adaptive filter, it is possible to realize a voice communication device that does not require an expensive camera device or complicated voice information processing. [Brief description of the drawings]

[0013] [Figure 1] FIG. 1 is a diagram showing a configuration of a voice communication device according to an embodiment of the present invention. [Diagram 2] FIG. 2 is a diagram showing an example of the arrangement of microphones and speakers. [Diagram 3] FIG. 3 is a block diagram showing the configuration of the beamformer unit. [Figure 4] FIG. 4 is a block diagram showing the configuration of the echo canceller. [Diagram 5] FIG. 5 illustrates an example of the configuration of the delay analysis unit. [Figure 6] FIG. 6 is a flow diagram showing the processing contents of the directivity analysis unit. [Figure 7]FIG. 7 is a diagram for explaining extraction of the inter-microphone characteristic filter coefficient. [Figure 8] FIG. 8 is a diagram showing an example of extracted inter-microphone characteristic filter coefficients. [Figure 9] FIG. 9 is a diagram showing an example of extracted inter-microphone characteristic filter coefficients. [Figure 10] FIG. 10 is a diagram showing an example of extracted inter-microphone characteristic filter coefficients. [Figure 11] FIG. 11 is a diagram illustrating an example of an estimation-oriented passband. [Figure 12] FIG. 12 is a diagram showing another example of an estimation-oriented passband. [Figure 13] FIG. 13 is a diagram showing an example of a passband of the direction of a speaker. [Figure 14] FIG. 14 is a diagram for explaining the weighting inter-microphone characteristic filter coefficient back conversion. [Figure 15] FIG. 15 is a flow chart showing the processing contents of the noise suppression unit. [Figure 16] FIG. 16 is a diagram for explaining the power calculation and maximum / minimum updating in the noise suppression unit. [Figure 17] FIG. 17 is a block diagram showing the configuration of the howling analysis unit. [Figure 18] FIG. 18 is a diagram for explaining the processing in the howling analysis unit. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0014] A voice communication device according to an embodiment of the present invention will be described below. FIG. 1 is a diagram showing a configuration example of a voice communication device 1. Note that each of the embodiments described below shows a preferred specific example of the present invention. The numerical values, components, arrangement positions and connection forms of the components, processing order in the flow charts, and the like shown in the following embodiment are merely examples and are not intended to limit the present invention. Also, each figure is not necessarily a strict illustration.

[0015] The voice communication device 1 of this embodiment is a voice communication device that is provided and connected to, for example, the front seat and the back seat of an automobile for hands-free communication.

[0016] The voice communication device 1 includes a beam former 10, an echo canceller 20, a microphone 11, a microphone 12, and a speaker 13. The microphones 11 and 12 are connected to the beam former 10, and the speaker 13 is connected to the echo canceller 20. The beam former 10 imparts directivity to the voice signals input from the microphones 11 and 12, passes the voice coming from a desired direction, and suppresses the voice coming from the other direction, forming a beamform. That is, the voice processing is performed to impart directivity so as to pass the voice of the speaker and block the voice from the other direction. The echo canceller 20 has an adaptive filter built-in, and estimates a pseudo echo signal of the input voice signal to suppress the echo, suppress noise, suppress howling, and distribute the loss of the transmission and reception.

[0017] 2 is a diagram showing the positional relationship between the microphones 11 and 12 connected to the beamformer unit 10 and the speaker 13. The center of the speaker 13 is located at the intersection of the equilateral sides of an isosceles triangle with the microphones 11 and 12, and the microphones 11 and 12 are located at the other intersections of the isosceles triangle. In this position, the sound of the speaker 13 can be treated as having equidistant directionality with respect to the two microphones.

[0018] 3 is a diagram showing the configuration of the beamformer unit 10. The beamformer unit 10 includes a delay analysis unit 111, a directivity analysis unit 112, a directional filter unit 113, a directional filtering subtraction unit 114, a reference microphone delay unit 115, and a delay analysis initial value storage unit 116. The input of the microphone 11 is input to the reference microphone delay unit 115, and the output of the delayed reference microphone delay unit 115 and the output of the microphone 12 are input to the delay analysis unit 111. The output of the delay analysis unit 111 is input to the directivity analysis unit 112, and the output of the directivity analysis unit 112 and the output of the microphone 12 are input to the directional filter unit 113. The output of the directional filter unit 113 and the output of the reference microphone delay unit 115 are input to the directional filtering subtraction unit 114, and the output of the directional filtering unit 115 is input to the directional filtering subtraction unit 114, and the output of the directional filtering unit 114 is input to the transmission input terminal of the echo canceller 20. Furthermore, a signal indicating the operating state of the adaptive filter of the echo canceller 20 (anti-loss analysis section state 260 ) is input to the delay analysis section 111 and the directivity analysis section 112 .

[0019] 4 is a diagram showing the configuration of the echo canceller 20. The echo canceller 20 has a sending input terminal 204, a sending output terminal 205, a receiving input terminal 202, and a receiving output terminal 203, and the speaker 13 is connected to the receiving output terminal 203. The sending input terminal 204 is connected to the sending output terminal of the beam former unit 10, and receives as input audio signals from the microphones 11 and 12 whose directivity has been controlled by the beam former unit 10.

[0020] The echo canceller 20 has a noise suppression section, an echo suppression section, a howling suppression section, and a loss amount distribution section, and suppresses echoes and background noise caused by the voice output from the speaker 13, and howling, and distributes the desired loss to the receiving path and the transmitting path so that a conversation can be established.

[0021] In the configuration of the echo canceller 20, the noise suppression section is the noise suppression section 212. The echo suppression section is implemented by the anti-loss analysis section 211, which is an adaptive filter, the environment delay section 213, the anti-loss filter section 214, the anti-loss subtraction section 215, and the anti-loss analysis initial value storage section 225. The howling suppression section is implemented by the howling analysis section 216 and the howling suppression filter section 217. The loss amount distribution section is implemented by the loss damper section 218, the transmitted voice determination section 219, the received voice determination section 220, the received loss insertion section 221, the transmitted loss insertion section 222, the transmitted side delay section 223, and the received side delay section 224. The operation of the echo canceller 20 will be described later.

[0022] (Explanation of the beamformer section) The operation of the beamformer unit 10 will be described. The beamformer unit 10 estimates an inter-microphone characteristic filter coefficient from the audio signals input from the two microphones 11 and 12 using an adaptive filter, converts the estimated inter-microphone characteristic filter coefficient into the frequency domain, and analyzes the phase difference between the audio from the two microphones from the phase of each frequency component. By grasping the sound to be passed from the phase difference analyzed in the frequency domain, a beamform is formed by imparting directivity so as to pass the audio in the desired direction and suppress the audio in the desired direction. The beamformer unit 10 is characterized in that it processes the audio signals from the microphones in the time domain without converting them into the frequency domain, and by analyzing the filter coefficients (filter coefficients representing the characteristics between the two microphones) obtained by the adaptive filter in the frequency domain, it obtains the phase difference between the sound sources arriving at the two microphones and imparts directivity.

[0023] The estimation by the adaptive filter of delay analysis unit 111 includes, as a filter coefficient, the characteristics with which one sound source is input to microphone 12, with the signal input to microphone 11 as a reference. That is, information on the phase difference, whether the audio signal of microphone 12 arrives early or late with respect to the audio of microphone 11, is included in estimated inter-microphone characteristic filter coefficient 152. By analyzing this inter-microphone characteristic filter coefficient 152, the phase difference with respect to the sound source can be obtained and directivity can be given.

[0024] The audio signal from microphone 11 is delayed by the length "mic1_delay" in reference microphone delay unit 115 and output. Based on the delayed audio signal from microphone 11 and the audio signal input from microphone 12, delay analysis unit 111 estimates the characteristics between the microphones as an inter-microphone characteristic filter coefficient using an adaptive filter, and outputs it as inter-microphone characteristic filter coefficient 152.

[0025] As shown in Fig. 5, the delay analysis unit 111 includes an ALP operating inter-microphone characteristic estimation adaptive filter 302 that operates while the anti-loss analysis unit 211, which is an adaptive filter of the echo canceller unit of the echo canceller 20, is operating in estimation, and a non-ALP operating inter-microphone characteristic estimation adaptive filter 303 that operates while the anti-loss analysis unit 211 is not operating. When the anti-loss analysis unit state 260 from the anti-loss analysis unit 211 is in adaptive estimation operation, the ALP operating inter-microphone characteristic estimation adaptive filter 302 performs estimation and outputs the inter-microphone characteristic filter coefficient 152. When the anti-loss analysis unit state 260 from the anti-loss analysis unit 211 is in non-adaptive estimation operation, the non-ALP operating inter-microphone characteristic estimation adaptive filter 303 performs estimation and outputs the inter-microphone characteristic filter coefficient 152. This makes it possible to reduce the influence of the voice output from the speaker 13 depending on whether the received voice is output from the speaker 13 or not. When a voice is being output from the speaker 13, the direction of the speaker cannot be detected, and if the adaptive filter of the delay analysis unit 111 is made to follow the sound when the speaker 13 is sounding, this will affect the convergence performance of the anti-loss analysis unit 211. For this reason, by providing an inter-microphone characteristic estimation adaptive filter for when the adaptive filter of the anti-loss analysis unit 211 of the echo canceller 20 is operating and when it is not operating, the effect on the convergence performance can be reduced, and when it is not operating, a coefficient can be generated that indicates the directivity of the speaker that is to be passed.

[0026] The length of the microphone characteristic filter coefficient is "mic1mic2_coeff_len." This corresponds to 192 taps when, for example, the sampling frequency of the digital signal processing is 16 kHz, the adaptive filter of the delay analysis unit 111 is an FIR filter, and the delay time is 12 ms.

[0027] The delay analysis initial value storage unit 116 stores the initial values ​​of the filter coefficients of the ALP-operating inter-microphone characteristic estimation adaptive filter 302, and uses them as the initial values ​​of the adaptive filter.

[0028] 6 is a diagram showing a flow of a directivity processing operation of the directivity analysis unit 112. The directivity analysis unit 112 inputs the inter-microphone characteristic filter coefficient 152 estimated by the delay analysis unit 111, converts it into the frequency domain, analyzes the phase difference from the inter-microphone characteristic filter coefficient in the converted frequency domain, and generates the directivity to be imparted between the two microphones as a weighting filter coefficient.

[0029] First, as shown in FIG. 7, a filter coefficient having the length of the inter-microphone characteristic filter coefficient is extracted from the beginning of the inter-microphone characteristic filter coefficient 152 (inter-microphone characteristic filter coefficient extraction 310). From the beginning of the inter-microphone characteristic filter coefficient 152, the length of "coeff_fft_len" is set as an inter-microphone characteristic frequency analysis input signal 401, with "mic1_delay offset" as the center. If the length "coeff_fft_len" is greater than "mic1mic2_coeff_len", which is the length of the inter-microphone characteristic filter coefficient 152, elements that are not overwritten are filled with "0", and if it is smaller, only the part where the conditions overlap is overwritten. This is to extract in a manner convenient for FFT (Fast Fourier Transform) so that frequency analysis can be performed from filter coefficients in which various characteristics are convoluted as characteristics between microphones, and a signal centered on "mic1_delay" is set as an input signal for FFT. Then, the inter-microphone characteristic frequency analysis input signal 401 is transformed into the frequency domain by FFT, and output as an inter-microphone characteristic complex filter coefficient 402 (inter-microphone characteristic FFT 311). Then, in phase difference calculation 312, the phase difference of the inter-microphone characteristic complex filter coefficients 402 (converted into the frequency domain) is obtained (phase difference calculation 312).

[0030] The calculation of the phase difference of the inter-microphone characteristic complex filter coefficients will be described. If we assume that the same signal magnitude is input to both microphones from the impulse response of the complex filter coefficients for inter-microphone characteristics, the phase difference when there is a one sample delay and one sample advance can be derived from the relationship between equations (3) to (5) below, leading to equation (6) which allows the angle to be calculated from time. This multiplies the frequency component F[n] in the direction you want to pass by "0 or a value close to 0" which is worthy of passing when it is included, and by "1 or a value close to 1" which is worthy of blocking when it is included in the direction you want to block. If this is done for the entire frequency domain and then returned to the time domain, the direction you want to pass will be removed and the direction you want to block will be attenuated, giving it directionality.

[0031] In the FFT, the Fourier transform [F(n)] can be expressed as the time domain sequence f(k) expressed by the relationship shown in equation (1).

number

number

number

[0032] In polar coordinates, the Fourier transform F[n] can be expressed as (Equation 4).

number

number

[0033] However, let 0 ≦ n < coeff_fft_len / 2 and diff[0] = 0. For diff[n], add or subtract 2π so that the difference from diff[n - 1] becomes the closest value. Since the value of diff[n] is affected by ambient noise and other speakers and has large fluctuations, the value after adjustment such as taking a moving average is used as the estimated steering value 403. Also, the directivity analysis 313 estimates the directivity from the following characteristics of the estimated steering value 403, and calculates an estimated steering value multiplier 404 for multiplying each sample value of the inter-microphone characteristic complex filter coefficient depending on whether it is a passband or a stopband.

[0034] Among the complex number sequences obtained by applying an FFT to a data sequence with a length of "coeff_fft_len" as shown in FIG. 8, where the offset "coeff_fft_len / 2" is "1" and the rest are "0", the phase difference between the complex number sequence [n] and the complex number sequence [n + 1] is considered to be π [rad] (Equation 6).

Equation

[0035] Also, among the complex number sequences obtained by applying an FFT to a data sequence with a length of "coeff_fft_len" as shown in FIG. 9, where the offset "coeff_fft_len / 2 + 1" is "1" and the rest are "0" from the beginning, the phase difference between the complex number sequence [0] and the complex number sequence "[coeff_fft_len / 2]" is considered to be π [rad] and have a linear phase (Equation 7).

Equation

[0036] Also, among the complex number sequences obtained by applying an FFT to a data sequence with a length of "coeff_fft_len" as shown in FIG. 10, where the offset "coeff_fft_len / 2 - 1" is "1" and the rest are "0" from the beginning, the phase difference between the complex number sequence [0] and the complex number sequence "[coeff_fft_len / 2]" is considered to be -π [rad] and have a linear phase (Equation 8).

number

[0037] The value of diff[n] at this time is the phase difference at which microphone 12 arrives one sample earlier than microphone 11 with respect to the sampling frequency. In light of this relationship, the threshold can be calculated from (Equation 9) with the delay time "delaytime" set as a specified value.

number

[0038] As shown in Fig. 11, assume that a speaker is present between the position of speaker n and the position of speaker f with respect to microphone 11 and microphone 12. In this case, at the position of speaker n, "Delay0near[s]" indicates that it takes longer for the voice to arrive at microphone 11 than it does at microphone 12, and at the position of speaker f, "Delay0far[s]" indicates that it takes longer for the voice to arrive at microphone 11 than it does at microphone 12. In this case, these are expressed as (Equation 10) and (Equation 11).

number

[0039] If the estimated directivity value 403 is in the range from "Taker0NearThres" to "Taker0FarThres", it is treated as a passband signal, and if it is outside this range, it is treated as a stopband signal. The shaded area is the area where passing is expected.

[0040] Also, as shown in FIG. 12, two passbands can be set by setting two delay values ​​for speaker 1 and speaker 2. At this time, it is expected that howling can be avoided by having the speaker 13 of the voice communication terminal of the other party in the stopband. Furthermore, even if it is a passband, each analysis frequency that is to be blocked is treated as a stopband. This mainly applies to the lowest analysis frequency. Furthermore, when the anti-loss analysis unit 211 is performing estimation operation in the two states of the anti-loss analysis unit state 260 of the echo canceller 20, the direction of the speaker 13 is treated as a speaker passband as shown in Fig. 13, and the speaker 13 passband is prioritized over the passband. This makes it possible to reduce the effect of the anti-loss analysis unit 211 on the adaptive filter.

[0041] The initial value of the estimated directivity value multiplier 404 is set to 1.0 (0 dB) based on the passband, stopband, and speaker passband, and the stopband is maintained (0 dB), a large attenuation amount (e.g., less than -40 dB) is applied in the passband, and an attenuation amount (e.g., -14 dB) that does not affect echo cancellation between the passband and stopband is applied in the speaker passband. Note that the positive and negative directions are reversed depending on the direction of the FFT rotor.

[0042] The directivity information addition unit 314 multiplies the inter-microphone characteristic complex filter coefficient 402 by an estimated directivity value multiplier 404 to output a weighted inter-microphone characteristic complex filter coefficient 405 . The inter-microphone characteristic IFFT 315 performs IFFT on the weighted inter-microphone characteristic complex filter coefficients 405 and outputs an inter-microphone characteristic frequency analysis output signal 406 .

[0043] 14, the inter-microphone characteristic filter coefficient back-conversion 316 overwrites the position of "coeff_fft_len / 2" of the inter-microphone characteristic frequency analysis output signal 406 to match the position of the offset "mic1_delay" of the weighting inter-microphone characteristic filter coefficient 151. If the length "mic1mic2_coeff_len" of the weighting inter-microphone characteristic filter coefficient 151 is greater than "coeff_fft_len", elements that have not been overwritten are filled with "0", and if it is smaller, only the parts where the conditions overlap are overwritten.

[0044] The directional filter unit 113 filters the audio signal from the microphone 12 with the weighted inter-microphone characteristic filter coefficient 151 , and the directional filtering subtraction unit 114 subtracts the filtered result from the output of the reference microphone delay unit 115 . Here, signals in frequency bands weighted by estimated directivity value multiplier 404 at less than 1.0 (0 dB) remain.

[0045] The above beamformer unit 10 has been described as performing beamforming based on the characteristics between the two microphones 11 and 12, but the number of microphones is not limited to two. When three microphones are provided, it is sufficient to use the sum of the delays of the two adjacent microphones.

[0046] (Explanation of echo canceller) The operation of the echo canceller 20 will now be described. For communication with a far-end call partner, a path from the receiving input terminal 202 through the receiving side delay unit 224 and the receiving loss insertion unit 221 to the receiving output terminal 203 is defined as a receiving path, and a path from the sending input terminal 204 through the noise suppression unit 212, the anti-loss subtractor 215, the howling suppression filter unit 217, the sending side delay unit 223, and the sending loss insertion unit 222 to the sending output terminal 205 is defined as a sending path. A speaker 13 is connected to the receiving output terminal 203, and a voice signal of the calling party is output to the speaker 13. Also, a voice signal of a near-end call partner is input to the sending input terminal 204 through the microphones 11 and 12 and the beamformer unit 10, enabling a hands-free conversation with the calling party at the far end.

[0047] The received voice determination unit 220 of the receiving path takes the power of the signal of the receiving input terminal 202, and when it exceeds a threshold value it determines that there is voice, and when it does not it determines that there is no voice. The receiving side delay unit 224 inserts a delay into the receiving path with a guideline of a time equivalent to the time it takes for the receiving voice determination unit 220 to produce a determination result. The receiving loss insertion unit 221 multiplies the output signal of the receiving side delay unit 224 by the receiving loss insertion amount 226 from the loss damper unit 218. The receiving output terminal 203 is connected to the output of the receiving side delay unit 224, and is connected to the speaker 13.

[0048] (Explanation of echo suppression section) The environment delay unit 213 outputs a delayed received output signal 253, which is delayed based on the arrival time from the received output terminal 203 to the noise-reduced sending path signal 251. The anti-loss analysis unit 211 estimates the relationship between the delayed received output signal 253 and the noise-reduced sending path signal 251 using an adaptive filter, and estimates a filter coefficient suitable for the anti-loss filter coefficient 254. That is, it performs an estimation operation of a pseudo echo signal as an echo canceller. When the anti-loss analysis unit 211 uses an estimation method such as LMS (Least Mean Square), it uses the autocorrelation of the delayed received output signal 253. When the value of the autocorrelation is equal to or greater than a certain value, it performs an estimation operation of the coefficient with the adaptive filter, and at that time, it outputs the anti-loss analysis unit status 260, which is a signal indicating the estimation state of the anti-loss analysis unit 211, as being in estimation operation. When the value is equal to or less than a certain value, it stops the estimation operation of the coefficient with the adaptive filter, and outputs the anti-loss analysis unit status 260 as being in non-estimation operation. The anti-loss analyzer state 260 is output to the beamformer unit 10 .

[0049] The anti-loss filter unit 214 filters the delayed received output signal 253 with an anti-loss filter coefficient 254 , and the anti-loss subtraction unit 215 subtracts the filtered delayed received signal from the noise-reduced transmit path signal 251 to output an anti-loss transmitted signal 259 .

[0050] The anti-loss signal 259 is input to the anti-loss analysis unit 211, and the amount of attenuation is monitored together with the noise-reduced transmission path signal 251. The anti-loss analysis unit 211 updates the anti-loss filter coefficients 254 when it is determined that more attenuation can be expected than with the current anti-loss filter coefficients 254. If the anti-loss analysis unit 211 can estimate an appropriate anti-loss filter coefficient 254, it is possible to attenuate the signal component output from the speaker 13 and included in the noise-reduced transmission path signal 251, and to reduce the amount of loss estimated by the loss damper unit 218. The anti-loss filter coefficient 254 that has been well attenuated is stored in the anti-loss analysis initial value storage unit 225 and can be used as the initial value at the next startup.

[0051] (Feedback suppression section) The howling suppression filter unit 217 filters the anti-loss signal 259 with a howling suppression filter coefficient 255 , and outputs a howling suppression filter unit output signal 256 .

[0052] (Explanation of loss distribution section) The transmitted voice determination unit 219 determines the presence or absence of voice from the noise-reduced transmission path signal 251 and the anti-loss signal 259. When the correlation between the noise-reduced transmission path signal 251 and the anti-loss signal 259 is high and the power of the noise-reduced transmission path signal 251 is high, or when the correlation between the noise-reduced transmission path signal 251 and the anti-loss signal 259 is low but the power of the noise-reduced transmission path signal 251 is high and the ratio of the power to the anti-loss signal 259 is small (less than 10 dB), it determines that the transmitted voice is present, and when the power of the noise-reduced transmission path signal 251 and the power of the anti-loss signal 259 are small or in other cases, it determines that the transmitted voice is not present.

[0053] The loss damper unit 218 distributes the amount of loss to the receiving loss insertion unit 221 and the sending loss insertion unit 222 . The area of ​​the anti-loss filter coefficient 254 is the reverberation amount, the result of dividing the anti-loss completed signal 259 by the noise-reduced transmission path signal 251 is the anti-loss attenuation amount, and the inverse of the result of multiplying the reverberation amount by the anti-loss attenuation amount is the terminal loss amount. Note that if the terminal loss amount exceeds 0 dB (1.0), it is set to 1. When the howling suppression filter unit output signal 256 does not have the signal size expected by the anti-loss filter, the loss amount to compensate for this is set to the NLP (Non Linear Processor) loss amount.

[0054] The loss damper unit 218 monitors the power of the signal at the receiving input terminal 202, and when it exceeds the receiving signal upper limit, holds the amount of loss required to reach the receiving signal upper limit as the receiving signal level loss. When the power of the signal at the receiving input terminal 202 becomes sufficiently small, the receiving signal level loss is amplified up to an upper limit of 0 dB. In addition, the loss damper unit 218 monitors the power of the signal at the howling suppression filter unit output signal 256, and when it exceeds the sending signal upper limit, holds the amount of attenuation required to reach the sending signal upper limit as the sending signal level loss. When the power of the howling suppression filter unit output signal 256 becomes sufficiently small, the sending signal level loss is amplified up to an upper limit of 0 dB. In addition, a silence loss is prepared as a loss during silence.

[0055] When the received voice determination unit 220 and the transmitted voice determination unit 219 determine that there is neither received voice nor transmitted voice, the reception path is determined to have the most attenuating loss among the end loss amount, the received signal level loss amount, or the silence loss amount as the reception loss insertion amount 226, and the transmission path is determined to have the most attenuating loss among the transmission signal level loss amount or the silence loss amount as the transmission loss insertion amount 227. When it is determined that there is no received voice but there is a transmitted voice, the most attenuating loss amount among the end loss amount, the received signal level loss amount, and the loss amount during silence is set as the received loss insertion amount 226 for the received path, and the transmitted signal level loss amount is set as the transmitted loss insertion amount 227 for the transmitted path. When it is determined that there is a received voice but not a transmitted voice, the received signal level loss amount is set as the received loss insertion amount 226 for the received path, and the most attenuating loss amount among the terminal loss amount, the NLP loss amount, and the transmitted signal level loss amount is set as the transmitted loss insertion amount 227 for the transmitted path. When it is determined that both the received voice and the transmitted voice are present, the loss amount which attenuates the most out of the square root of the end loss amount or the received signal level loss amount is set as the received loss insertion amount 226 for the received path, and the loss amount which attenuates the most out of the square root of the end loss amount or the transmitted signal level loss amount is set as the transmitted loss insertion amount 227 for the transmitted path, and so on. In this way, the loss amounts are appropriately distributed to the received loss insertion unit 221 and the transmitted loss insertion unit 222 according to the voice determination.

[0056] (Description of noise suppression section) 15, the noise suppression unit 212 receives a transmission voice from the transmission input terminal 204, performs noise reduction processing, and outputs a noise-reduced transmission path signal 251. The noise suppression unit 212 of the echo canceller 20 of this embodiment performs noise reduction processing in a predetermined frequency width in the frequency domain before the echo suppression unit that uses an adaptive filter, and outputs the noise-reduced transmission path signal 251.

[0057] FIG. 15 is a flow chart illustrating the process of the noise suppression unit 212. First, the voice signal input to the sending input terminal 204 is subjected to FFT and converted to the frequency domain (sending path FFT 502). The sending path FFT 502 consecutively arranges "noise_fblength / 2" pieces of data from the sending input terminal 204 and "noise_fblength / 2" pieces of data from one frame before, composes "noise_fblength" pieces of data, multiplies them by a Hanning window, performs FFT, and outputs a sending input complex signal 602 of length "noise_fblength / 2+1". The length of "noise_fblength" is selected so that one sampling point has a width of 100 Hz to 200 Hz. Then, in power calculation 503, the sending input complex signal 602 is applied to (Equation 4), the magnitude ``|F(n)|'' is set to the sending input spectrum 653, and the sending input power 603 obtained by averaging the sending input spectrum 653 is output.

[0058] In the maximum / minimum update 504, subbands as shown in FIG. 16 are formed, and the minimum value of the transmission input power 603 and its position as a sample point are stored and updated. FIG. 16 shows an example of subbands, where "noise_fblength" is divided into four regions, and region "0" has two subbands, with one data item of the sending input power 603 as one subband, and region "1" has two subbands, with three data items of the sending input complex signal 602 as one subband. The number of data items is defined for each region, and each subband is configured with the length "noise_sblen." The minimum power value and its position are stored in a minimum power history storage unit 604 for each subband.

[0059] An example of obtaining the minimum value of each sub-band is shown. Let the transmission input power 603 be sin_power[m], and the minimum power history holding unit 604 be powmin[bf][sbm]. bf is the length of the time to be held with the multiplicity of powmin. sbm satisfies 0 ≦ sbm < noise_sblen. In FIG. 16, when obtaining the minimum value of sub-band 3, if the minimum value among sin_power[5], sin_power[6], sin_power[7], and powmin[updateidx][3].pow is sin_power[5], then update powmin[updateidx][3].pow with sin_power[5], and update powmin[updateidx][3].pos with 5 as the position storage. Here, 0 ≦ updateidx < bf. If powmin[updateidx][3].pow is the minimum value, it is held. The same operation is performed for all sub-bands. This is done for powmin[updateidx] for a certain period, for example, 100 ms. When updateidx reaches noise_sblen while incrementing by 1, it is set to "0" and repeated. Thus, the minimum value and its position over a long time (for example, 1 second to 2 seconds) are held. For the maximum power history holding unit 654, the maximum value and its position are also obtained in the same way, and the position is held in the maximum power history holding unit 654.

[0060] The sub-band noise level estimation 505 takes the smallest minimum power value for each sub-band among the values held in the minimum power history holding unit 604 as the noise level, and averages the minimum power value and the position at that time with time constants respectively, and holds and outputs them to the sub-band noise level estimation holding unit 605. Also, when it is desired to increase the attenuation amount of the noise, the maximum power is selected from the maximum power history holding unit 654 of each sub-band. When the difference between the maximum power and the minimum power is large, a value that has little influence on the estimation operation of the anti-loss analysis unit 211, for example, ((maximum power) - (minimum power)) × 0.1, etc., is added to the sub-band noise level estimation holding unit 605.

[0061] The subband noise level estimation full band extension 506 complements the minimum value and its position of the subband noise level estimation storage unit 605 from the minimum values ​​and positions of adjacent subbands, and outputs a full band estimated noise level 606 extended to the full band length “noise_fblength / 2+1”. The fullband level estimation correction 507 performs correction for each sample point of the fullband estimated noise level 606 in response to changes in the sending input spectrum 653, and updates the fullband estimated noise level 606 that is greater than the sending input spectrum 653 with the sending input spectrum 653, and outputs a corrected estimated noise complex signal 607 in which the fullband estimated noise level 606 is set to |F[n]| in (Equation 4) and θ is set to the same value as the sending input complex signal 602.

[0062] The noise level subtraction 508 outputs a noise-reduced send input complex signal 608 obtained by subtracting the corrected estimated noise complex signal 607 from the send input complex signal 602 .

[0063] The output spectrum peak hold 509 monitors the spectrum of the noise-reduced sending input complex signal 608 for each sample point, and updates the corresponding value. If the spectrum is larger than the sending input peak hold 252, the value of the sending input peak hold 252 is updated. If the spectrum does not fall below a value obtained by multiplying the sending input peak hold 252 by a value less than 1.0, for example, 0.5, for a first fixed time (for example, 50 ms) or more, the noise suppression instantaneous howling suppression attenuation value 257 for the corresponding sample point is successively multiplied by a value less than 1.0, for example, 0.001 dB, until the spectrum starts to attenuate or the limit value of the noise suppression instantaneous howling suppression attenuation value 257 is reached, for example, to -6 dB. If the spectrum does not start to attenuate for a second fixed time, for example, 200 ms or more, the sending input peak hold state is established. This sending input peak hold state is released when the hold release 261 is turned on. When the spectrum value is sufficiently small rather than in the transmission input peak hold state, the noise suppression instantaneous howling suppression attenuation value 257 is successively multiplied by a value greater than 1.0, such as 0.001 dB, until it becomes 1.0 or greater, and then the multiplication is stopped and the value is set to 1.0. After the noise suppression instantaneous howling suppression attenuation value 257 becomes 1.0, the value of the transmission input peak hold 252 is multiplied by a value less than 1.0, such as −0.001 dB, until it becomes the current spectrum, and when it reaches that value, the multiplication is stopped.

[0064] The send path IFFT 510 performs IFFT (Inverse Fast Fourier Transform) on the noise-reduced send input complex signal 608, multiplies it by a Hanning window, and combines it with the result of the previous frame processing to output a noise-reduced send path signal 251 of length “noise_fblength / 2”.

[0065] (Explanation of howling suppression section) The howling suppression section performs howling suppression on the anti-loss processed transmission signal 259 which has been subjected to subtraction and output by the anti-loss subtraction section 215 . 17, the howling analysis unit 216 multiplies the attenuation amounts output from the base howling suppression analysis unit 801 and the instantaneous howling suppression analysis unit 802 in a howling suppression filter synthesis unit 803 to generate a howling suppression filter coefficient 255. The howling suppression filter coefficient 255 has a length of about 1 ms to 2 ms.

[0066] The base howling suppression analysis unit 801 updates the base howling suppression attenuation value 901 when the anti-loss filter coefficient 254 is updated. A length including a sample point that is a peak of the coefficient is extracted from the anti-loss filter coefficient 254 as shown in Fig. 18, and converted to a frequency domain by FFT, focusing on the spectrum, attenuating sample points below a certain level (for example, -20 dB from the peak) relative to the spectrum peak as shown in Fig. 18, and outputting a base howling suppression attenuation value 901 that matches the length of the howling suppression filter coefficient 255 that attenuates sample points that show a spectrum above the peak of the sending input peak hold 252 and the upper limit value of the normal signal. When the base howling suppression attenuation value 901 is updated, the hold release 261 is turned ON, and if the sending input peak hold state is released in the noise suppression section, the hold release 261 is turned OFF and output.

[0067] The instantaneous howling suppression analysis unit 802 outputs an instantaneous howling suppression attenuation value 902 obtained by adjusting the noise suppression instantaneous howling suppression attenuation value 257 to the length of the howling suppression filter coefficient 255 . The howling suppression filter synthesis unit 803 compares the base howling suppression attenuation value 901 and the instantaneous howling suppression attenuation value 902 for each sample point, selects the smaller value, applies IFFT to it, and outputs it as the howling suppression filter coefficient 255.

[0068] The voice communication device of this embodiment can be placed in the front and rear seats of a car and connected to enable hands-free conversation within the same space of the vehicle. In addition, since it suppresses surrounding noise, it can also suppress noise while driving. Although the voice communication device of this embodiment has been described as a voice communication device for use in a car, it can also be used as a device for making calls in a noisy environment, for example, within the same space such as a factory or warehouse.

[0069] The voice communication device 1 as an embodiment has been described above, but the beam former unit 10 and the echo canceller 20 of this voice communication device 1 can execute each process of the beam former unit 10 and the echo canceller 20 by installing a software program in a general-purpose information processing device equipped with a CPU and memory.

[0070] [Summary of the invention] As described above, the voice communication device 1 includes microphones 11 and 12, microphone 11 is used as a reference microphone, and a beamformer unit 10 that imparts directivity to voice input to microphones 11 and 12 by passing voice arriving from a desired direction and suppressing voice arriving from other directions. The beamformer unit 10 includes an adaptive filter that estimates an inter-microphone characteristic filter coefficient indicating the characteristics of the voice input to microphone 12 compared to the voice input to microphone 11, frequency-converts the estimated inter-microphone characteristic filter coefficient, detects the phase difference between the voice signals from microphones 11 and 12 from the converted inter-microphone characteristic filter coefficient, and imparts directionality to the voice signals from microphones 11 and 12 based on the phase difference.

[0071] This allows the phase difference between the microphones to be detected by the adaptive filter and directivity to be imparted to the audio signal, thereby realizing a voice communication device that does not require expensive camera devices or complex image information processing.

[0072] The beamformer unit 10 includes a delay analysis unit 111 which is an adaptive filter that estimates an inter-microphone characteristic filter coefficient 152 using the audio signals input from the microphones 11 and 12, a directivity analysis unit 112 which converts the estimated inter-microphone characteristic filter coefficient 152 into the frequency domain, detects the phase difference between the audio signals input from the microphones 11 and 12, and determines the directional direction of the passband or stopband of the audio input to the microphones 11 and 12, a directional filter unit 113 which passes audio in the pass direction determined by the directivity analysis unit 112, generates a filter coefficient that imparts directivity to block audio in the block direction, converts the generated filter coefficient into the time domain, and weights the audio signal from the microphone 12 with the converted filter coefficient, and a subtraction unit 114 which subtracts the audio signal from the microphone 12 that has been weighted by the directional filter unit 113 from the audio signal input from the microphone 11, and passes the audio signal in the pass direction.

[0073] This allows beamforming using adaptive filters.

[0074] The voice communication device 1 is provided with an echo canceller 20 that outputs an audio signal input to a sending input terminal 204 from a sending output terminal 205, outputs an audio signal input to a receiving input terminal 202 from a speaker 13 of a receiving output terminal 203, and is provided between the sending input terminal 204 and the sending output terminal 205 to suppress an echo generated by the audio signal output from the receiving output terminal 203, and the echo canceller 20 performs frequency conversion on the audio signal input to the sending input terminal 204 to detect noise in a predetermined frequency range, and suppresses noise components for each predetermined frequency range. The system includes a suppression unit 212, an anti-loss analysis unit 211 which is an echo suppression unit that has an adaptive filter for estimating a pseudo echo signal due to a received audio signal output from the speaker 13 of the received output terminal 203 and inputting the audio signal from the sending input terminal 204 into the audio signal in which the noise components have been suppressed by the noise suppression unit 212, an environment delay unit 213, an anti-loss filter unit 214, an anti-loss subtraction unit 215, and a howling analysis unit 216 and a howling suppression filter unit 217 which are howling suppression units that perform howling suppression processing on the echo-suppressed audio signal.

[0075] This allows clear voice communication even in noisy surroundings.

[0076] The voice communication device 1 includes a beam former unit 10 and an echo canceller 20.

[0077] This enables clear, hands-free voice communication even in the same noisy space.

[0078] The delay analysis unit 111 includes an ALP operating inter-microphone characteristic estimation adaptive filter 302 which performs adaptive operation while the adaptive filter of the anti-loss analysis unit 211 is in estimation operation, and a non-ALP operating inter-microphone characteristic estimation adaptive filter 303 which performs adaptive operation when the estimation operation is not in progress, and operates the operating inter-microphone characteristic estimation adaptive filter 302 when an audio signal of the receiving input terminal 202 is output to the receiving output terminal 203 and the adaptive filter of the anti-loss analysis unit 211 is in estimation operation, and operates the non-operating inter-microphone characteristic estimation adaptive filter 303 when no audio signal is output to the receiving output terminal.

[0079] This allows the adaptive filter of the anti-loss analysis unit 211 to estimate an appropriate anti-loss filter coefficient for attenuating the signal component output from the receiving output terminal. [Explanation of symbols]

[0080] 1 Voice communication device 10 Beamformer section 11,12 Mike 13 Speaker 20 Echo Canceller 111 Delay Analysis Unit 112 Directivity analysis section 113 Directional filter section 114 Directional filtering subtraction section 115 Reference microphone delay section 116 Delay analysis initial value memory unit 202 Receiving input terminal 203 Receiver output terminal 204 Transmitting input terminal 205 Transmitting output terminal 211 Anti-loss Analysis Department 212 Noise suppression section 213 Environmental Delay Division 214 Anti-loss filter section 215 Anti-loss subtraction section 216 Feedback Analysis Section 217 Howling suppression filter section 218 Loss Damper Part 219 Transmitted voice judgment unit 220 Received voice judgment unit 221 Receiver loss insert 222 Transmission loss insertion part 223 Transmitter Delay Unit 224 Receiver Delay Unit 225 Anti-loss analysis initial value memory section 302 Adaptive filter for estimating microphone characteristics during ALP operation 303 Adaptive Filter for Estimating Microphone Characteristics When Non-ALP is in Operation 801 Bass feedback suppression analysis section 802 Real-time feedback suppression analysis section 803 Howling suppression filter synthesis section

Claims

1. A voice communication device including a plurality of microphones, one of which is a reference microphone, and a beam former that imparts directivity to a voice input to the plurality of microphones so as to pass a voice coming from a desired direction and suppress a voice coming from another direction, The beam former unit includes: a delay analysis unit which is an adaptive filter that estimates an inter-microphone characteristic filter coefficient using the audio signals input from the plurality of microphones; a directivity analysis unit that converts the estimated inter-microphone characteristic filter coefficient into a frequency domain, detects a phase difference between the audio signals input from the multiple microphones, and determines a directivity direction of a passband or a stopband of the audio input to the multiple microphones; a directional filter unit that generates a filter coefficient that provides directivity for passing the sound in the passing direction determined by the directional analysis unit and blocking the sound in the blocking direction, converts the filter coefficient into a time domain, and weights the sound signals from microphones other than the reference microphone with the converted filter coefficient; a subtraction unit that subtracts the audio signals from the other microphones that have been weighted by the directional filter unit from the audio signal input from the reference microphone, and passes the audio signal in the passing direction; A voice communication device comprising:

2. A voice signal input to the transmission input terminal is output from the transmission output terminal, and a voice signal input to the reception input terminal is output from a speaker of the reception output terminal; an echo canceller provided between the sending input terminal and the sending output terminal for suppressing an echo generated by the voice signal output from the receiving output terminal; A voice communication device comprising: The echo canceller comprises: a noise suppression unit that performs frequency conversion on the voice signal input to the transmission input terminal, detects noise in a region of a predetermined frequency width, and suppresses noise components for each predetermined frequency width; an echo suppression unit having an adaptive filter for estimating a pseudo echo signal based on a received voice signal output from a speaker of the received voice output terminal, and suppressing an echo from the voice signal in which a noise component has been suppressed by the noise suppression unit; a feedback suppression unit that performs feedback suppression processing on the echo-suppressed audio signal; A voice communication device comprising:

3. The output of the beamformer unit according to claim 1 is connected to the transmission input terminal of the voice communication device according to claim 2. A voice communication device comprising:

4. 4. The voice communication device according to claim 3, The delay analysis unit the adaptive filter of the echo suppression unit includes an operating inter-microphone characteristic estimation adaptive filter that performs an adaptive operation during an estimation operation, and a non-operating inter-microphone characteristic estimation adaptive filter that performs an adaptive operation when the estimation operation is not performed, When the voice signal of the receiving input terminal is output to the receiving output terminal and the adaptive filter of the echo suppression unit is performing an estimation operation, the operating inter-microphone characteristic estimation adaptive filter is operated, and when the voice signal is not output to the receiving output terminal, the non-operating inter-microphone characteristic estimation adaptive filter is operated. A voice communication device comprising:

5. 5. The voice communication device according to claim 4, the echo suppression unit has a storage unit that stores the estimated filter coefficient that can suppress the echo as an initial value; The delay analysis unit has a storage unit that stores the filter coefficients estimated by the operating microphone characteristic estimation filter as initial values. A voice communication device comprising:

6. On the computer, A beamformer unit including a plurality of microphones, one of the plurality of microphones being a reference microphone, and imparting directivity to sounds input to the plurality of microphones such that a sound coming from a desired direction passes through and a sound coming from another direction is suppressed, a delay analysis unit which is an adaptive filter that estimates an inter-microphone characteristic filter coefficient using the audio signals input from the plurality of microphones; a directivity analysis unit that converts the estimated inter-microphone characteristic filter coefficient into a frequency domain, detects a phase difference between the audio signals input from the plurality of microphones, and determines a directivity direction of a passband or a stopband of the audio input to the plurality of microphones; a directional filter unit that generates a filter coefficient that provides directivity for passing the sound in the passing direction determined by the directional analysis unit and blocking the sound in the blocking direction, converts the filter coefficient into a time domain, and weights the sound signals from microphones other than the reference microphone with the converted filter coefficient; a subtraction unit that subtracts the audio signal from another microphone that has been weighted by the directional filter unit from the audio signal input from the reference microphone, and passes the audio signal in the passing direction; A program that executes the processing.

7. To your computer A voice signal input to the transmission input terminal is output from the transmission output terminal, and a voice signal input to the reception input terminal is output from a speaker of the reception output terminal; an echo canceller provided between the sending input terminal and the sending output terminal to suppress an echo generated by the voice signal output from the receiving output terminal; A voice communication device comprising: The echo canceller, A voice signal input to the transmission input terminal is frequency-converted to detect noise in a predetermined frequency range, and noise components in each predetermined frequency range are suppressed; an adaptive filter for estimating a pseudo echo signal based on a received voice signal output from a speaker at the receiving output terminal, thereby suppressing echo from the voice signal in which noise components have been suppressed; Performs feedback suppression on echo-suppressed audio signals A program that executes a process.

Images