TW202234386A - Method, communication device and communication system for double-talk detection and echo cancellation - Google Patents

Abstract

A method for double-talk detection and echo cancellation adapted to a communication device having a voice processing circuit, wherein the communication device is configured to electrically connect to a speaker and a microphone, and the method comprises: obtaining a far-end signal by the communication device, wherein the far-end signal is configured to be played by the speaker, generating a recording signal by the microphone, performing, by the voice processing circuit, a voice processing procedure according to the far-end signal and the recording signal to generate a near-end amplitude, generating, by the voice processing circuit, a determination result according to the near-end amplitude, a historic data and a threshold data at least, wherein the determination result is configured to indicate a condition of double-talk condition or single-talk, and adjusting a parameter set of a dynamic equalizer according to the determination result by the voice processing circuit.

Description

Method, communication device and communication system for detecting bilateral speech and eliminating echo

The invention relates to a hands-free device in a remote audio or video conference, in particular to a method, a communication device and a communication system for detecting bilateral speech and eliminating echoes.

In order to meet the needs of large-scale conferences, the hands-free devices used in video conferences are usually connected in series with multiple hands-free devices, the speaker output volume is increased, and a high-sensitivity microphone is used to provide an environment that is convenient for participants to listen to and receive audio from the hands-free device.

However, the above approach will cause the microphone of the hands-free device to receive more echoes, so that it is difficult to achieve full duplex when executing the echo cancellation algorithm.

In view of this, the present invention provides a method, a communication device and a communication system for detecting bilateral speech and eliminating echoes, thereby solving the problem that the echo cancellation algorithm is difficult to achieve full duplex when the microphone receives too many echoes.

According to an embodiment of the present invention, the method for detecting bilateral speech and eliminating echoes is suitable for a communication device. The communication device has an audio processing circuit, and the communication device is used for electrically connecting a speaker and a microphone. The method includes: the communication device obtains a remote signal, The far-end signal is used for speaker playback; the microphone produces a recording; the audio processing circuit performs an audio processing procedure according to the far-end signal and the recording to generate the near-end amplitude; the audio-processing circuit generates a judgment result based on at least the near-end amplitude, historical information and threshold information , the judgment result is used to indicate bilateral speech or unilateral speech; and the audio processing circuit adjusts the parameter group of the dynamic equalizer according to the judgment result.

According to an embodiment of the present invention, a communication device for detecting bilateral calls and eliminating echoes includes: a communication circuit for obtaining a remote signal; and an audio processing circuit for electrically connecting the communication circuit and for electrically connecting a speaker and a microphone, The audio processing circuit is used for transmitting the far-end signal to the speaker, receiving the recording produced by the microphone, and performing the audio processing procedure according to the far-end signal and the recording to generate the near-end amplitude, at least according to the near-end amplitude, historical information and threshold information for indicating The judgment result of bilateral speech or unilateral speech, and the parameter group of the dynamic equalizer is adjusted according to the judgment result, wherein the dynamic equalizer processes the far-end signal according to the parameter group.

According to an embodiment of the present invention, a communication system for detecting bilateral speech and eliminating echoes includes a first communication device and a second communication device that are communicatively connected to each other, wherein the first communication device includes: a first communication interface for obtaining a remote terminal signal; the first audio processing circuit is electrically connected to the first communication interface and used to electrically connect the first speaker and the first microphone, the first audio processing circuit is used to transmit the remote signal to the first speaker and receive the first microphone The generated first recording generates a first judgment result according to the remote signal, the first recording and the second judgment result. The first judgment result is used to indicate bilateral speech or unilateral speech, and the dynamic equalizer is adjusted according to the first judgment result. a parameter group, wherein the dynamic equalizer processes the remote signal according to the parameter group; and a first communication circuit is electrically connected to the first audio processing circuit and used for sending the remote signal processed by the dynamic equalizer and receiving the first communication circuit. Two judgment results; the second communication device includes: a second communication circuit for receiving the remote signal sent by the first communication device; a second audio processing circuit for electrically connecting to the second communication circuit and for electrically connecting to the second speaker and the second microphone, the second audio processing circuit is used to transmit the remote signal to the second speaker, receive the second recording generated by the second microphone, and generate a second judgment result according to the remote signal and the second recording. It is used for instructing bilateral speech or unilateral speech; wherein the second communication circuit is further used for sending the second judgment result to the first communication device.

The above description of the present disclosure and the following description of the embodiments are used to demonstrate and explain the spirit and principle of the present invention, and provide further explanation of the scope of the patent application of the present invention.

The detailed features and characteristics of the present invention are described in detail in the following embodiments, and the content is sufficient to enable any person skilled in the relevant art to understand the technical content of the present invention and implement accordingly, and according to the content disclosed in this specification, the scope of the patent application and the drawings , any person skilled in the related art can easily understand the related concepts and features of the present invention. The following examples further illustrate the viewpoints of the present invention in detail, but do not limit the scope of the present invention in any viewpoint.

The applicable scene of the present invention is, for example, playing audio or video through one or more online conference speakerphones in a large conference room.

FIG. 1 is a hardware architecture diagram of a communication system for detecting double-talking and eliminating echoes according to an embodiment of the present invention. Bilateral speaking is defined as: while the near-end (local) speaker plays the far-end participant's voice, the near-end microphone records the near-end participant's voice.

In the embodiment shown in FIG. 1 , the communication system includes a first communication device 10 , a second communication device 20 and a third communication device 30 that are communicatively connected to each other. The three communication devices 10 , 20 and 30 are connected in series in a daisy chain. In practice, the number of communication devices in the communication system can be determined according to the size of the conference space. If only one communication device is required, the first communication device 10 is set. If more than two communication devices are required, one first communication device 10 and more than one second communication device 20 are provided. The third communication device 30 in FIG. 1 has the same hardware structure as the second communication device 20 . Therefore, only the implementation details of the first and second communication devices 10 and 20 are described below.

The first communication device 10 includes a first communication interface 11 , a first communication circuit 12 , a first audio processing circuit 14 and a first conversion circuit 16 . The first communication interface 11 is connected to the network N for communication. The first audio processing circuit 14 is electrically connected to the first communication interface, and is electrically connected to the first communication circuit 12 , the first speaker 18 and the first microphone 19 through the first conversion circuit 16 .

The first communication interface 11 receives the remote signal from the remote device from the network N. The far-end signal contains the voice information of the far-end participants.

The first communication circuit 12 connects the first and second communication devices 10 and 20 in series. The first communication circuit 12 can send the remote signal processed by the first audio processing circuit 14 and the first converting circuit 16 and receive the second judgment result from the second communication device 20 . The first communication circuit 12 adopts, for example, a Low Voltage Differential Signaling (LVDS) technology, and uses differential signals to transmit data, thereby increasing the transmission distance.

The first audio processing circuit 14 can transmit the remote signal to the first speaker 18, receive the first recording generated by the first microphone 19, and generate a first judgment result according to the remote signal, the first recording and the second judgment result. The result is used to indicate bilateral speech or unilateral speech, and according to the first judgment result, the parameter group of the dynamic equalizer (Dynamic Equalizer) is adjusted, wherein the dynamic equalizer processes the far-end signal according to the parameter group. The first audio processing circuit 14 uses, for example, a central processing unit (Central Processing Unit, CPU). The detection method of bilateral speech will be separately described later.

The first conversion circuit 16 can concentrate the data and convert the data format. The first conversion circuit 16 is, for example, a field programmable gate array (Field Programmable Gate Array, FPGA). In other embodiments, the first conversion circuit 16 can also be integrated into the first audio processing circuit 14 .

The first speaker 18 can play the far-end signal adjusted by the dynamic equalizer according to the parameter set adjusted by the first audio processing circuit 14 to generate the first far-end sound. The first microphone 19 can record the sound of the environment in which it is located to generate the first recording. In practice, the first communication device 10 can have a built-in or externally connected first speaker 18 and a first microphone 19 .

The second communication device 20 includes a second communication circuit 22 , a second audio processing circuit 24 and a second conversion circuit 26 . The second communication circuit 22 is communicatively connected to the first communication circuit 12 . The second audio processing circuit 24 is electrically connected to the second communication circuit 22 , the second speaker 28 and the second microphone 29 through the second conversion circuit 26 .

The second communication circuit 22 can receive the remote signal sent by the first communication device 10 , and send the second judgment result of the second audio processing circuit 24 for the bilateral call to the first communication device 10 . The second communication circuit 22 is basically the same as the first communication circuit 12 in hardware.

The second audio circuit 24 can transmit the remote signal to the second speaker 28, receive the second recording generated by the second microphone 29, and generate a second judgment result according to the remote signal and the second recording. The second judgment result can indicate bilateral speaking or One-sided speech.

The second converting circuit 26 is electrically connected to the second audio processing circuit 24 . The second conversion circuit 26 is substantially the same as the first conversion circuit 16 in hardware.

The second speaker 28 can play the remote signal received from the first communication device 10 to generate the second remote sound. The second microphone 29 can record the sound of the environment in which it is located to generate a second recording. In practice, the second communication device 20 can be built in or externally connected to the second speaker 28 and the second microphone 29 .

FIG. 2 is a schematic diagram of the internal modules of the first and second audio circuits 14 and 24 according to an embodiment of the present invention. The blocks in FIG. 2 can be implemented in software or hardware circuits. The arrows between the squares are used to indicate the direction of data transfer. For the convenience of inspection, the first communication interface 13 , the first and second communication circuits 12 and 22 and the first and second conversion circuits 16 and 26 are omitted in FIG. 2 .

FIG. 3 is a flowchart of a method for detecting double-talking and eliminating echoes according to an embodiment of the present invention, which is applicable to the first or second communication device 10 or 20 shown in FIG. 1 . Please refer to Figure 2 and Figure 3 together below.

The step S31 is "the communication device obtains the remote signal", and the step S32 is "the microphone generates a recording".

First, the first communication interface 11 obtains the remote signal from the network N, and the dynamic equalizer 141 processes the remote signal according to the preset parameter set, and then outputs the remote signal to the first and second speakers 18 and 28 . The first speaker 18 plays the first far-end sound according to the processed far-end signal. At this time, the first microphone 19 records the sound around the first communication device 10 to generate the first recording; A recording will contain the participant's voice and the first far-end voice. Similarly, the second speaker 28 plays the second far-end sound and the second microphone 29 generates the second recording.

Step S33 is "the audio processing circuit performs an audio processing procedure according to the far-end signal and the recording to generate at least the near-end amplitude".

Please refer to FIG. 2, the first recording goes through an adaptive filter (Adaptive Filter) 143, a nonlinear processor (NLP) 144, a noise suppressor (Noise Reduction, NR) 145, a mixer 146, automatic gain The processing of modules such as a controller (Automatic Gain Control, AGC) 147 and a compressor (Compressor) 148 is finally transmitted to a remote communication device through the network N. In the above module, the adaptive filter 143 and the nonlinear processor 144 can perform Acoustic Echo Cancelling (AEC), the noise suppressor 145 can suppress noise, and the mixer 146 combines the two noise suppressors The respective output signals of 145 and 245 are mixed. The automatic gain controller 147 adjusts the loudness in the output signal of the mixer 146, and the compressor 148 avoids sound saturation.

The modules such as the adaptive filter 243 , the nonlinear processor 244 and the noise suppressor 245 in the second audio circuit 24 operate as described above. The difference is that the noise suppressor in the second audio circuit 24 The output signal of 245 is further serially connected to the mixer 146 in the first audio circuit 14 .

The double-talk detectors 142 and 242 in the first and second audio circuits 14 and 24 are respectively used to detect whether there is a double-talk situation around the first and second communication devices 10 and 20 .

Among multiple communication devices connected in a daisy chain, the communication device at the end will make a comprehensive judgment based on the near-end amplitude, far-end speech probability, far-end amplitude, historical information, and threshold information to generate a judgment result of the bilateral call. Near-end amplitude reflects the volume of the near-end sound. Far-end speech probability reflects the probability that the far-end voice is speech. The far-end amplitude reflects the volume of the far-end sound. The historical information includes the near-end amplitude attenuation value, the far-end speech probability attenuation value and the far-end amplitude attenuation value, and the three attenuation values reflect the previous near-end amplitude, far-end speech probability and far-end amplitude. The threshold information includes an interval formed by a high threshold value and a low threshold value, and the high threshold value is greater than the low threshold value. In addition to relying on the aforementioned various information, the communication device not located at the end also refers to another judgment result of the next-level communication device to generate the judgment result of the bilateral call. Taking FIG. 1 as an example, the subsequent communication device of the first communication device 10 is the second communication device 20 , and the subsequent communication device of the second communication device 20 is the third communication device 30 .

Step S34 is "the audio processing circuit generates a judgment result for indicating bilateral speech or unilateral speech at least according to the near-end amplitude, history information and threshold information." Assuming that the communication system of an embodiment of the present invention only includes the second communication device 20 located at the end and the first communication device 10 not located at the end, two implementations for generating the judgment result are illustrated below.

In the first embodiment of step S34, the bilateral speech detector 242 in the second communication device 20 performs the following operations:

First, the audio frame (frame) output by the noise suppressor 245 this time is divided into several blocks in units of K sampling points. The maximum amplitude is obtained for the K sampling points in each block, and then the maximum amplitude values of all blocks are averaged. The average amplitude obtained as above is the near-end amplitude A2.

Next, the near-end amplitude attenuation value A2' is obtained from the historical information, which is the product of the previous calculation result of the noise suppressor 245 and the attenuation coefficient. The previous first flag F21' is obtained from the historical information, which is the result of the previous setting of the second audio circuit 24. The high threshold value TH and the low threshold value TL are obtained from the threshold information, where TH>TL. The history information and the threshold information are stored, for example, in the storage unit of the second audio circuit 10 .

Next, select the larger A2 _max of the near-end amplitude A2 and the near-end amplitude attenuation value A2'.

If TH<A2 _max , set the first flag F21 to 1;

If A2 _max <TL, set the first flag F21 to 0;

If _TL≤A2 max≤TH, the set value of the first flag F21 is maintained, that is, the previous first flag F21' is used as the current first flag F21; and

Multiply A2 _max by the attenuation coefficient as the near-end amplitude attenuation value A2' for the next calculation.

When the first flag F21 is set (F21=1), the second judgment result indicates double-sided speech. When the first flag F21 is released (F21=0), the second judgment result indicates that the unilateral speech is made.

Please refer to the aforementioned operation flow. The noise suppressor 145 in the first communication device 10 generates the near-end amplitude A1 in the aforementioned manner. The first audio processing circuit 14 obtains the near-end amplitude attenuation value A1' and the previous first flag F11' from the historical information, obtains the high threshold value TH and the low threshold value TL from the threshold information, and then sets and cancels the above-mentioned method. Or maintain the first flag F11. When the first flag F21 is set (F21=1), the first judgment result indicates double-sided speaking. The difference from the above is that when the first flag F11 in the first communication device 10 is released (F11=0) but the first flag F22 in the second communication device 20 is set (F21=1), the first The judgment result still indicates bilateral speaking. Unless the first flags F11 and F21 of the two communication devices 10 and 20 are both released, the first judgment result will indicate a unilateral call. In short, the bilateral speech detector 142 not only performs the operations of the bilateral speech detector 242, but also receives the second judgment result through the first communication circuit 12 as a reference.

In the second embodiment of step S34, the second audio processing circuit 24 performs an audio processing procedure according to the far-end signal and the second recording to generate the near-end amplitude A2, the far-end speech probability P2 and the far-end amplitude F2.

The historical information obtained by the bilateral speech detector 242 includes the near-end amplitude attenuation value A2', the far-end speech probability attenuation value P2', the far-end amplitude attenuation value F2', the previous first flag F21', and the previous second flag Flag F22' and the previous third flag F33'. The threshold information obtained by the bilateral speech detector 242 includes a near-end amplitude interval R1, a far-end speech interval R2 and an end amplitude interval R3, wherein each interval Rn is composed of a high threshold value THn and a low threshold value TLn, that is, Rn=[ THn, TLn], and THn>TLn, n∈{1, 2, 3}.

The operation process for the bilateral speech detector 242 to generate the second judgment result according to at least the near-end amplitude A2, the historical information and the threshold information is as follows:

Select the first larger A2 _max of the near-end amplitude A2 and the near-end amplitude A2' attenuation value;

When the first larger A2 _max is greater than the near-end amplitude interval R1, that is, TH1 < A2 _max , the first flag F21 is set to 1;

When the first larger A2 _max is smaller than the near-end amplitude interval R1, that is, A2 _max <TL1, the first flag F21 is released to 0;

When the first larger A2 _max is within the near-end amplitude range R1, that is, TL1 < A2 _max < TH1, the previous first flag F21' is used as the current first flag F21;

selecting the second larger P2 _max in the far-end speech probability P2 and the far-end speech probability attenuation value P2';

When the second larger P2 _max is greater than the far-end speech interval R2, that is, TH2 < P2 _max , the second flag F22 is set to 1;

When the second larger P2 _max is smaller than the far-end speech interval R2, that is, P2 _max <TL2, the second flag F22 is released to 0;

When the second larger P2 _max is within the far-end speech interval R2, that is, TL2 < A2 _max < TH2, the previous second flag F22' is used as the current second flag F22;

Select the third larger F2 _max of the far-end amplitude F2 and the far-end amplitude attenuation value F2';

When the third larger F2 _max is greater than the remote amplitude interval R3, that is, TH3 < F2 _max , the third flag F23 is set to 1;

When the third larger F2 _max is smaller than the remote amplitude interval R3, that is, F2 _max <TL3, the third flag F23 is released to 0;

When the third larger F2 _max is within the far-end amplitude range R3, that is, TL3 < F2 _max < TH3, the previous second flag F23' is used as the current second flag F23; and

Multiply the first, second and third larger A2 _max , P2 _max and F2 _max by the corresponding attenuation coefficient to be used as the near-end amplitude attenuation value A2' and the far-end speech probability attenuation value required in the next calculation P2' and the far-end amplitude attenuation value F2'.

After the bilateral speech detector 242 completes the above calculation process, the following judgment logic is adopted according to the three flags F21, F22 and F23: when the first flag F21 is set, or when the second flag F22 and the third flag are set When the flags F23 are all set, the second judgment result D2 is indicated as bilateral speech, that is, D2=(F21 or (F22 and F23)). If D2 is 1, it means that the second communication device 20 is currently talking on both sides; if D2 is 0, it means that the second communication device 20 is currently talking on one side.

In addition, the second communication circuit 22 further transmits the second determination result D2 to the double-talk detector 142 of the first communication device 10 .

In the same way, compared to the calculation method of the bilateral speech detector 242, the bilateral speech detector 142 in the first audio processing circuit 14 is based on the near-end amplitude A1, the far-end speech probability P1, the far-end amplitude F1 and the data corresponding to the three data. The history information and the threshold information of the three flags F11 , F12 and F13 belonging to the first communication device 10 can be calculated. It should be noted that: the judgment logic of the first judgment result D1 further refers to the second judgment result D2, that is, D1=(F11 or (F12 and F13) or D2). If D1 is 1, it means that the first communication device 10 is currently speaking bilaterally; if D1 is 0, it means that the first communication device 10 is currently speaking unilaterally.

In the second implementation of step S34, if one of the following three conditions is met, the bilateral speech detectors 142 and 242 determine that the speech is bilateral:

(1) The volume of the most recent subsampled near-end sound exceeds a threshold;

(2) The probability that the most recently played far-end sound belongs to speech exceeds another threshold and the volume of the far-end sound exceeds another threshold; and

(3) The latter-level communication device is judged to be bilaterally speaking (the communication device belonging to the end of the daisy chain does not need to consider this situation).

Step S35 is "the audio processing circuit adjusts the parameters of the dynamic equalizer according to the judgment result".

After the first determination result D1 is generated, the first audio processing circuit 14 adjusts the parameter group of the dynamic equalizer according to the first determination result D1.

In one embodiment, the parameter group of the dynamic equalizer can be adjusted to a unilateral parameter group or a bilateral parameter group, and each parameter group includes one or more parameters. As the default value when the first communication device 10 is powered on, the unilateral parameter group is used for pure voice or pure echo. The bilateral parameter group is used in the case of bilateral speech.

FIG. 4 is a flow chart of the first audio processing circuit 14 for setting dynamic equalizer parameters.

Step S41 is "starting the detection of bilateral speech", the details of which are as described above.

The flow of steps S42 to S43 represents that when the judgment result indicates bilateral speaking, it is judged whether the parameter group of the current dynamic equalizer 141 is a bilateral parameter group. When the determination in step S43 is "Yes", step S44 is executed to maintain the current setting of the bilateral parameter group. When the determination in step S43 is “No”, step S45 is executed to gradually adjust the parameter set of the dynamic equalizer 141 from the unilateral parameter set to the bilateral parameter set until the curve of the dynamic equalizer 141 conforms to the setting of the bilateral parameter set .

The flow of steps S42 to S46 represents that when the judgment result indicates that unilateral speaking is indicated, it is judged whether the parameter group of the current dynamic equalizer 141 is a unilateral parameter group. When the determination in step S46 is "Yes", step S47 is executed, and the current setting of the one-sided parameter group is maintained. When the determination in step S46 is "No", step S48 is executed to gradually adjust the parameter set of the dynamic equalizer 141 from the bilateral parameter set to the unilateral parameter set until the curve of the dynamic equalizer 141 conforms to the setting of the unilateral parameter set until.

After completing steps S44, S45, S46 and S47, the process returns to step S41.

One embodiment of the "gradual adjustment" described in steps S45 and 4S8 is: for each parameter that can be set in the dynamic equalizer, adjust a parameter in the parameter group according to an offset every unit time, until This parameter is equal to the one-side/two-side parameter in the one-side/two-side parameter group. The gain value of a unilateral parameter in the unilateral parameter group is greater than the gain value of a bilateral parameter in the bilateral parameter group. The parameters in the bilateral parameter group are set at negative gain, thereby suppressing the output volume of the first and second speakers 18 and 28 . The parameter values in the bilateral parameter group are strongly suppressed at frequencies with severe speaker distortion, and weaker or even no suppression at other frequencies with less severe distortion. Therefore, the distortion degree of the far-end speech in the sound recorded by the first and second microphones 19 and 29 can be reduced.

To sum up, the method, communication device and communication system for detecting bilateral speech and eliminating echoes proposed by the present invention can not only improve the audio reception and listening distance of online conferences, but also can perform echo cancellation function in full duplex mode. The present invention can improve the fluency of the online conference, and bring all the participants a sense of presence in the same place. The present invention can avoid the situation that when one party speaks, the other party's voice cannot be transmitted. By adjusting the parameters of the dynamic equalizer, the present invention can suppress the frequency band of the speaker's nonlinear distortion, and in other frequency bands, the adjustment range can be reduced or even not adjusted. Therefore, the serial speakerphone used in the present invention does not need to increase the volume of a single speaker, so it is avoided that in a large conference room, in order to increase the volume of the speaker, the participants close to the speaker are exposed to excessively loud speakers. Discomfort. In addition, the serial hands-free device used in the present invention does not need to be equipped with a high-sensitivity microphone.

Although the present invention is disclosed in the foregoing embodiments, it is not intended to limit the present invention. Changes and modifications made without departing from the spirit and scope of the present invention belong to the scope of patent protection of the present invention. For the protection scope defined by the present invention, please refer to the attached patent application scope.

N: network 10: The first communication device 11: The first communication interface 12: The first communication circuit 14: The first audio processing circuit 16: The first conversion circuit 18: The first horn 19: First Mic 20: Second communication device 22: The second communication circuit 24: Second audio processing circuit 28: Second speaker 29: Second Microphone 30: Third communication device 32: The third communication circuit 34: The third audio processing circuit 38: Third Horn 39: Third Microphone 141: Dynamic Equalizer 142, 242: Bilateral speech detector 143, 243: Adaptive filter 144, 244: Nonlinear processors 145, 245: Noise suppressor 146: Mixer 147: Automatic Gain Controller 148: Compressor S31~S35, S41~S49: Steps

1 is a hardware architecture diagram of a communication system for detecting bilateral speech and eliminating echoes according to an embodiment of the present invention; 2 is a schematic diagram of the internal modules of the first and second audio circuits according to an embodiment of the present invention; FIG. 3 is a flowchart of a method for detecting double-sided speech and eliminating echoes according to an embodiment of the present invention; and FIG. 4 is a flow chart of the first audio processing circuit for setting the parameter group of the dynamic equalizer.

S31~S35: Steps

Claims (9)

A method for detecting bilateral speech and eliminating echo, which is suitable for a communication device, the communication device has an audio processing circuit, and the communication device is used for electrically connecting a speaker and a microphone, and the method includes: obtain a remote signal from the communication device, the remote signal is used for playback by the speaker; produce a sound recording with the microphone; using the audio processing circuit to perform an audio processing procedure according to the far-end signal and the recording to generate a near-end amplitude; generating a judgment result by the audio processing circuit at least according to the near-end amplitude, a history information and a threshold information, and the judgment result is used to indicate bilateral speech or unilateral speech; and A parameter group of the dynamic equalizer is adjusted by the audio processing circuit according to the judgment result. The method of claim 1, wherein The historical information includes a near-end vibration attenuation value and a previous judgment result; the threshold information includes a high threshold value and a low threshold value, and the high threshold value is greater than the low threshold value; and Generating the judgment result based on at least the near-end amplitude, the historical information and the threshold information includes: select the larger of the near-end amplitude and the near-end amplitude attenuation value; When the larger one is greater than the high threshold value, indicating that the judgment result is bilateral speaking; When the larger one is less than the low threshold value, selectively indicating that the judgment result is unilateral speaking; and When the larger one is between the high threshold value and the low threshold value, the previous determination result is used as the determination result. As the method of claim 1, it further includes: The audio processing circuit is used to perform the audio processing procedure according to the far-end signal and the recording to generate a far-end speech probability and a far-end amplitude; wherein The historical information includes a near-end amplitude attenuation value, a far-end speech probability attenuation value, a far-end amplitude attenuation value, a previous first flag, a previous second flag and a previous third flag; the threshold information includes a near-end amplitude interval, a far-end speech interval and a far-end amplitude interval; Generating the judgment result based on at least the near-end amplitude, the historical information and the threshold information includes: selecting a first larger of the near-end amplitude and the near-end amplitude attenuation value; When the first larger one is greater than the near-end amplitude interval, setting a first flag; When the first larger one is smaller than the near-end amplitude interval, release the first flag; and When the first larger one is within the near-end amplitude range, use the previous first flag as the first flag; selecting a second larger of the far-end speech probability and the far-end speech probability attenuation value; When the second larger one is greater than the far-end voice interval, setting a second flag; When the second larger one is smaller than the far-end speech interval, release the second flag; and When the second larger one is within the far-end voice interval, use the previous second flag as the second flag; selecting a third larger of the far-end amplitude and the far-end amplitude attenuation value; When the third larger one is greater than the far-end amplitude interval, setting a third flag; When the third larger one is smaller than the far-end amplitude interval, release the third flag; and When the third larger one is within the far-end amplitude interval, use the previous third flag as the third flag; and When the first flag is set, or when both the second flag and the third flag are set, it indicates that the judgment result is double-talk. The method of claim 3, further comprising: using the communication device to obtain another determination result from another communication device that is communicatively connected to the communication device, and Generating the judgment result based on at least the near-end amplitude, the historical information and the threshold information further includes: When the other judgment result indicates double-talk, it indicates that the judgment result is double-talk. As claimed in the method of claim 1, adjusting the parameter group of the dynamic equalizer by the audio processing circuit according to the judgment result includes: When the judgment result indicates unilateral speaking, judge whether the parameter group is a unilateral parameter group; When the parameter group is the unilateral parameter group, maintain the parameter group; and When the parameter group is not the unilateral parameter group, adjust the parameter group to the unilateral parameter group; and When the judgment result indicates bilateral speaking, judge whether the parameter group is a bilateral parameter group; when the parameter set is the bilateral parameter set, maintain the parameter set; and When the parameter group is not the bilateral parameter group, the parameter group is adjusted to the bilateral parameter group, wherein The gain value of a unilateral parameter in the unilateral parameter group is greater than the gain value of a bilateral parameter in the bilateral parameter group. As in the method of claim 2, where Adjusting the parameter group to the unilateral parameter group includes: adjusting a parameter in the parameter group according to an offset every unit time until the parameter in the parameter group is equal to the unilateral parameter; and Adjusting the parameter group to the bilateral parameter group includes: adjusting the parameter in the parameter group according to the offset every unit time until the parameter in the parameter group is equal to the bilateral parameter. A communication device for detecting bilateral speech and eliminating echo, comprising: a communication circuit for obtaining a remote signal; and an audio processing circuit electrically connected to the communication circuit and used for electrically connecting a speaker and a microphone, wherein the audio processing circuit is used for transmitting the remote signal to the speaker, receiving a recording generated by the microphone, and according to the remote The end signal and the recording are subjected to an audio processing procedure to generate a near-end amplitude, at least according to the near-end amplitude, a history information and a threshold information to generate a judgment result for indicating bilateral speech or unilateral speech, and according to the judgment As a result, a parameter set of a dynamic equalizer is adjusted, wherein the dynamic equalizer processes the remote signal according to the parameter set. The communication device of claim 7, wherein the communication circuit is further configured to obtain another judgment result of another communication device, and the other judgment result is used to indicate that the other communication device is bilateral or unilateral, and the audio The processing circuit further updates the judgment result according to the other judgment result. A communication system for detecting bilateral calls and eliminating echoes, comprising a first communication device and a second communication device that are communicatively connected to each other, wherein The first communication device includes: a first communication interface for obtaining a remote signal; a first audio processing circuit electrically connected to the first communication interface and used for electrically connecting a first speaker and a first microphone, the first audio processing circuit is used for transmitting the remote signal to the first speaker, Receive a first recording generated by the first microphone, generate a first determination result according to the remote signal, the first recording and a second determination result, and the first determination result is used to indicate bilateral speech or unilateral speech, and adjusting a parameter set of a dynamic equalizer according to the first judgment result, wherein the dynamic equalizer processes the remote signal according to the parameter set; and a first communication circuit electrically connected to the first communication circuit and used for sending the remote signal processed by the dynamic equalizer and receiving the second judgment result; The second communication device includes: a second communication circuit for receiving the remote signal sent by the first communication device; a second audio processing circuit electrically connected to the second communication circuit and used for electrically connecting a second speaker and a second microphone, the second audio processing circuit is used for transmitting the remote signal to the second speaker, Receive a second recording generated by the second microphone, and generate the second judgment result according to the remote signal and the second recording, and the second judgment result is used to indicate bilateral speech or unilateral speech; wherein The second communication circuit is further used for sending the second judgment result to the first communication device.

Images