Echo cancellation processing method and device in scene with large loudspeaker loudness

Abstract

The invention discloses an echo cancellation processing method and device in a scene with large loudspeaker loudness, and belongs to the field of voice signal processing. The echo cancellation processing device is installed in equipment and comprises an audio acquisition module, a self-adaptive threshold compression module, a delay determination module, a delay compensation module, an echo cancellation module and an audio playing module. The echo cancellation method comprises the following steps: when equipment runs on line, the audio acquisition module acquires a reference signal sent by far-end equipment and a near-end microphone signal, and inputs the reference signal and the near-end microphone signal into the adaptive threshold compression module to compress the amplitude of the microphone signal; meanwhile, according to the delay tau stored in advance, delay compensation is carried out on the far-end reference signal through a delay compensation module; and finally, performing block frequency domain adaptive filtering on the signal subjected to the compression amplitude limiting processing and the far-end reference signal subjected to the compensation delay through an echo cancellation module, and performing echo cancellation. According to the method, the convergence speed of the linear filter is higher and more stable.

Description

Technical Field

[0001] The present invention belongs to the field of speech signal processing, and in particular to an echo cancellation processing method and device in a scene with high loudness of a speaker. Background Art

[0002] Outdoor communication equipment needs to have a long sound propagation distance and sound pickup distance, so the loudspeaker must be loud and the microphone must be sensitive. If there is a lack of physical isolation measures inside the communication equipment, there will be strong echo coupling inside the equipment cavity, and the energy intensity of the echo signal collected by the equipment will far exceed the energy intensity of the reference signal.

[0003] In this scenario, the strength of the echo signal collected by the microphone differs greatly from the strength of the far-end reference signal. The adaptive linear echo cancellation filter of the existing traditional echo cancellation algorithm is difficult to converge, resulting in poor overall echo cancellation performance.

[0004] Traditional algorithms do not consider compressing the signals collected by the microphone that exceed a certain threshold. Therefore, based on the actual application scenario, the present invention introduces an adaptive compression scheme based on the traditional algorithm, that is, adaptively compressing the microphone collected signal before linear echo cancellation.

[0005] Since the delay between the microphone acquisition signal and the reference signal of the hard echo device is usually small and fixed, the delay can be determined when the device is manufactured and written into the medium that will not be lost when the power is off; at the same time, according to the strength difference between the microphone acquisition signal and the reference signal, the amplitude of the microphone acquisition signal is adaptively adjusted, which can effectively improve the convergence ability of the linear echo cancellation adaptive filter, thereby improving the overall echo cancellation effect of the device. Summary of the invention

[0006] The present invention provides an echo cancellation processing method and device in a scene with loud speaker volume. On the basis of a traditional echo cancellation algorithm, adaptive compression preprocessing of a microphone collected signal is added. In a scene with loud speaker volume and high microphone sensitivity, the convergence ability of an adaptive filter in a linear echo cancellation stage of a traditional algorithm can be improved, thereby improving the overall echo cancellation effect of the device.

[0007] The echo cancellation processing device is installed inside the device, and includes an audio acquisition module, an adaptive threshold compression module, a delay determination module, a delay compensation module, an echo cancellation module and an audio playback module;

[0008] First, the locally stored audio file is played, and the audio acquisition module collects the echo signal, which is input into the delay determination module, and the delay τ is output and written into the storage medium that is not lost when the power is off;

[0009] Then, when the device is online, the audio acquisition module collects the far-end reference signal and the near-end speaker signal sent by the far-end device, and inputs them into the adaptive threshold compression module for compression. At the same time, according to the delay τ stored in advance, the far-end reference signal is delayed compensated by the delay compensation module;

[0010] Finally, the echo cancellation module performs echo cancellation on the far-end reference signal and the reference signal after the delay compensation.

[0011] The echo cancellation processing method comprises the following steps:

[0012] Step 1: For the calling device A that is in a call, the audio acquisition module collects the remote reference signal and near-end microphone signal

[0013] N represents the number of sample points of the signal, which is set according to engineering experience;

[0014] Step 2: The adaptive threshold compression module adaptively compresses the amplitude of the microphone signal d in the time domain to obtain the signal

[0015] Adaptive compression includes the following processing steps:

[0016] Step 201: Calculate the RMS value x of the remote reference signal x rms , and converted to intensity decibel value x dB Right now

[0017]

[0018] x=[x0,x1,…,x i ,…,x N-1 ], x[i] represents the i-th element in the remote reference signal x;

[0019] Step 202: Determine the intensity decibel value x dB Is it greater than the threshold T1? If yes, execute step 203; otherwise, execute step 206;

[0020] T1 represents the threshold for determining whether there is a speaker signal at the far end, which is preset based on engineering experience;

[0021] Step 203, compress the near-end microphone signal d to obtain a compressed near-end microphone signal d′;

[0022] The specific compression includes the following steps:

[0023] Step I: according to the strength decibel value d of the near-end microphone signal dB , combined with the preset compression threshold T2 and compression ratio R, the gain g is calculated:

[0024]

[0025] The compression threshold T2 and the compression ratio R are preset based on engineering experience;

[0026] Step II, using the gain g to calculate the compressed near-end microphone collected signal d′,

[0027]

[0028] Where d[i] represents the i-th element in the near-end microphone signal d, where 0≤i≤N-1;

[0029] Step 204: performing a limiting process on the compressed near-end microphone collected signal d′, the specific formula is:

[0030]

[0031] Wherein, 0≤i<N, T3 is the amplitude limiting threshold, and α is the amplitude attenuation coefficient; T3 and α are preset according to engineering experience.

[0032] Step 205: using the signal d′ collected by the near-end microphone after limiting, the compression threshold T2 is updated as the next compression threshold;

[0033] The specific process is:

[0034] First, the RMS value d′ is calculated using the signal d′ collected by the near-end microphone after limiting. rms , and converted into intensity decibel value d′ dB ;

[0035] Then, calculate the decibel value x of the remote reference signal dB The intensity decibel value d′ of the signal collected by the near-end microphone after limiting dB The signal strength difference between them is e;

[0036] As shown below: e = d' dB -x dB ;

[0037] Next, when the signal strength difference e is greater than the threshold T4, the near-end microphone is used to collect the signal strength decibel value d′ after limiting. dB Update compression threshold T2:

[0038] The update formula is as follows: T2 = T2-(d′ dB -x dB )×μ;

[0039] T4 is the threshold for determining signal strength difference, which is preset based on engineering experience; μ is the update step length, which is preset based on engineering experience;

[0040] Step 206: Do not perform any processing on the signal collected by the near-end microphone, that is, the compressed near-end microphone signal d′ is the near-end microphone signal d;

[0041] d′=d.

[0042] Step 3: Device A runs a delay determination module to obtain a delay τ value of its own device and compensates for a remote reference signal x;

[0043] The delay compensation is performed only on the remote reference signal x containing N sample points in the first frame, specifically:

[0044] According to the delay τ pre-set in device A and the remote reference signal sampling rate f, the number of delayed sample points is calculated:

[0045] K = f × τ;

[0046] Add K zeros to the front end of the remote reference signal x, that is,

[0047] Take the first N sample points of the far-end reference signal after the front-end is padded with 0, that is,

[0048] The remaining K sample points are used as the first K sample points of the far-end reference signal containing N sample points in the next frame;

[0049] Set the delay τ for device A. The specific steps are as follows:

[0050] First, in offline state, the speaker of device A plays the local audio file, and the microphone collects the signal And perform adaptive compression processing to obtain the compressed microphone acquisition signal

[0051] Then, the local audio file is used as the reference signal According to the preset candidate delay set {τ1,τ2,,τ m}, and perform delay compensation on the reference signal to obtain {r1,r2,,r m};

[0052] Where m is the number of candidate delays, delay τ i Corresponding to the compensation signal r i , i∈[1,m];

[0053] Next, the compressed signal s′ and the compensated signal r i , respectively, perform frequency domain adaptive filtering in each block segment, and obtain the linear filtering result {y1,y2,,y m};

[0054] Finally, according to the linear filtering result {y1,y2,,y m} and the compressed signal s′, respectively calculate the echo return loss enhancement and obtain {l1,l2,,l m}; According to the set {l1,l2,,l m}, the index i corresponding to the maximum value in the candidate delay set {τ1,τ2,,τ m}, i.e., τ = τ i As the delay of device A's setup.

[0055] Step 4: performing block-segment frequency domain adaptive filtering according to the signal d′ after compression and limiting processing and the far-end reference signal x after delay compensation;

[0056] The advantages of the present invention are:

[0057] Compared with the classic block-segment frequency domain adaptive filtering method, the present invention makes the linear filter converge faster and more stable in the scenario where the echo signal strength collected by the microphone is significantly different from the far-end reference signal strength, thereby achieving better echo cancellation effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 This is a structural diagram of an echo cancellation processing device for a loud speaker scene according to the present invention;

[0059] Figure 2 It is a flow chart of an echo cancellation processing method in a scenario with high speaker loudness according to the present invention;

[0060] Figure 3 This is a schematic diagram of a remote reference signal in a scenario with high speaker loudness according to the present invention;

[0061] Figure 4 This is a schematic diagram of a signal collected by a near-end microphone in a scenario with a loud speaker;

[0062] Figure 5 It is a schematic diagram of a speaker after adaptive threshold compression processing in a scene with high loudness according to the present invention;

[0063] Figure 6 This is a schematic diagram of the echo cancellation result in a scenario with a loud speaker; DETAILED DESCRIPTION

[0064] In order to facilitate those skilled in the art to understand and implement the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. Obviously, the described embodiments are only partial embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work should fall within the scope of protection of the present invention.

[0065] The echo cancellation processing device is installed inside the device, such as Figure 1 As shown, it includes an audio acquisition module, an adaptive threshold compression module, a delay determination module, a delay compensation module, an echo cancellation module and an audio playback module;

[0066] The audio acquisition module is used to pick up the target speaker and the echo signal;

[0067] The adaptive threshold compression module is used to compress the signal collected by the near-end microphone;

[0068] The delay determination module obtains the delay between the near-end microphone acquisition signal and the far-end reference signal;

[0069] The audio playback module is used to play the remote reference signal and the audio file stored locally;

[0070] The adaptive echo cancellation module is used to perform echo cancellation on the compressed microphone acquisition signal and the far-end reference signal.

[0071] First, the locally stored audio file is played through the audio playback module; at the same time, the echo signal collected by the audio acquisition module is input into the delay determination module, the delay τ is output, and the delay τ is written to the storage medium that is not lost when the power is off;

[0072] Then, when the device is online, the audio acquisition module collects the far-end reference signal and the near-end speaker signal sent by the far-end device, and then inputs them into the adaptive threshold compression module for compression. At the same time, according to the delay τ stored in advance, the far-end reference signal is delayed by the delay compensation module;

[0073] Finally, the echo cancellation module performs echo cancellation on the far-end reference signal and the reference signal after the delay compensation.

[0074] The echo cancellation processing method is as follows: Figure 2 As shown, the following steps are included:

[0075] Step 1: For the calling device A that is in a call, the audio acquisition module collects the remote reference signal and near-end microphone signal

[0076] N represents the number of sample points of the signal, which is set according to engineering experience;

[0077] The near-end microphone collected signal is a digital signal obtained after analog-to-digital conversion of the analog signal collected by the microphone of the calling device A, such as Figure 4As shown; the far-end reference signal is the digital signal obtained by the near-end device after the calling device A receives the signal sent by the called device B, such as Figure 3 As shown;

[0078] like Figure 3 and 4 As shown, the signal collected by the microphone of the calling device A during double talk includes the echo and the speaker signal of the calling device A; when the near end is single talk, it only includes the speaker signal of the calling device A; the collected echo signal has been clipped, and its strength is much higher than the far-end reference signal strength, and much higher than the speaker signal collected by the microphone of the calling device A;

[0079] Step 2: The adaptive threshold compression module adaptively compresses the amplitude of the microphone signal d in the time domain to obtain the signal

[0080] Adaptive compression includes the following processing steps:

[0081] Step 201: Calculate the RMS value x of the remote reference signal x rms , and converted to intensity decibel value x dB Right now

[0082]

[0083] x=[x0,x1,,x i ,,x N-1 ], x[i] represents the i-th element in the remote reference signal x;

[0084] Step 202: Determine the intensity decibel value x dB Is it greater than the threshold T1? If yes, execute step 203; otherwise, execute step 206;

[0085] T1 represents the threshold for determining whether there is a speaker signal at the far end, which is preset based on engineering experience;

[0086] Step 203, compress the near-end microphone signal d to obtain a compressed near-end microphone signal d′;

[0087] The specific compression includes the following steps:

[0088] First, according to the strength decibel value d of the near-end microphone signal dB , combined with the preset compression threshold T2 and compression ratio R, the gain g is calculated:

[0089]

[0090] The compression threshold T2 and the compression ratio R are preset based on engineering experience;

[0091] Then, the compressed near-end microphone signal d′ is calculated.

[0092]

[0093] Where d[i] represents the i-th element in the near-end microphone signal d, where 0≤i≤N-1;

[0094] Step 204: performing a limiting process on the compressed near-end microphone collected signal d′, the specific formula is:

[0095]

[0096] Wherein, 0≤i<N, T3 is the amplitude limiting threshold, and α is the amplitude attenuation coefficient; T3 and α are preset according to engineering experience.

[0097] The schematic diagram after the adaptive threshold compression and limiting processing in this example is as follows Figure 5 As shown in the figure. After the compression and limiting processing, the strength difference between the echo signal collected by the microphone of the calling device A and the speaker signal is reduced, providing favorable conditions for the subsequent execution of block-segment frequency domain adaptive filtering;

[0098] Step 205: using the signal d′ collected by the near-end microphone after limiting, the compression threshold T2 is updated as the next compression threshold;

[0099] The specific process is:

[0100] Calculate the RMS value d′ of the near-end microphone collected signal d′ after limiting rms , and converted into intensity decibel value d′ dB ;

[0101] Then, calculate the decibel value x of the remote reference signal dB The intensity decibel value d′ of the signal collected by the near-end microphone after limiting dB The signal strength difference between them is e;

[0102] As shown below: e = d' dB -x dB ;

[0103] Next, when the signal strength difference e is greater than the threshold value T4, the near-end microphone is used to collect the signal strength decibel value d′ after limiting. dB Update compression threshold T2;

[0104] The update formula is as follows: T2 = T2-(d′ dB -x dB )×μ;

[0105] Among them, μ is the update step size, which is preset according to engineering experience; T4 is the threshold for determining signal strength difference, which is preset according to engineering experience;

[0106] Step 206: Do not process the signal collected by the near-end microphone, that is, d′=d;

[0107] Step 3: Device A runs a delay determination module to obtain a delay τ value of its own device and compensates for a remote reference signal x;

[0108] The delay determination module is used to determine the delay (unit: ms) between the near-end microphone acquisition signal obtained by the hard echo device and the far-end reference signal; when the device performs the echo cancellation function, the delay determination module is first run to obtain the delay τ, and then the echo cancellation is performed;

[0109] It is worth noting that when delay compensation is performed on the remote reference signal x, delay compensation is performed only on the remote reference signal x containing N sample points in the first frame, specifically:

[0110] According to the delay τ pre-set in device A and the remote reference signal sampling rate f, the number of delayed sample points is calculated:

[0111] K = f × τ;

[0112] Add K zeros to the front end of the remote reference signal x, that is,

[0113] Take the first N sample points of the far-end reference signal after the front-end is padded with 0, that is,

[0114] The remaining K sample points are used as the first K sample points of the far-end reference signal containing N sample points in the next frame;

[0115] Set the delay τ for device A. The specific steps are as follows:

[0116] First, the microphone acquisition signal is the signal collected after the microphone is turned on synchronously when the device speaker plays the local audio file;

[0117] When the device runs the delay determination module, the read thread and the write thread are enabled;

[0118] The reading thread is used to obtain the data collected by the microphone from the sound card buffer;

[0119] The write thread is used to write the data of the local audio file into the sound card buffer;

[0120] After the device determines the delay for the first time, it writes the delay τ to the corresponding storage medium that is not lost when power is off;

[0121] Collect the microphone signal of device A in offline state Perform adaptive compression processing to obtain the microphone acquisition signal after compression processing

[0122] The signal s is the signal collected by the microphone when the speaker of device A plays a local audio file;

[0123] Then, the local audio file is used as a reference signal According to the preset candidate delay set {τ1,τ2,,τ m}, and perform delay compensation on the reference signal to obtain {r1,r2,,r m};

[0124] Where m is the number of candidate delays, delay τ i Corresponding to the compensation signal r i , i∈[1,m];

[0125] Next, the compressed signal s′ and the compensated signal r i , perform block frequency domain adaptive filtering (PBFDAF, Partitioned Block Frequency Domain Adaptive Filter) respectively, and obtain the linear filtering result {y1,y2,,y m};

[0126] Finally, according to the linear filtering result {y1,y2,,y m} and the compressed microphone signal s′, respectively calculate the echo return loss enhancement (EREL, Echo Return Loss Enhancement), and get {l1,l2,,l m}; An indicator used to judge whether the echo cancellation effect is good or bad;

[0127] According to the set {l1,l2,,l m}, the index i corresponding to the maximum value in the candidate delay set {τ1,τ2,,τ m}, i.e., τ = τ i As the delay of the setting of device A; where 1≤i≤m;.

[0128] Step 4: Based on the signal d′ after compression and limiting processing and the far-end reference signal x after delay compensation, perform block-segment frequency domain adaptive filtering to eliminate echo.

[0129] The result diagram of echo cancellation in this example is as follows: Figure 6 The method adopted by the present invention has no effect on the microphone signal collected by device A when the device A is speaking alone; in the scenario where the echo signal strength collected by the microphone of device A is much greater than the speaker signal strength, the echo can be effectively suppressed when the far end is speaking alone or with two people.

[0130] It should be noted that those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the implementation methods of the present invention, and it should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed in the present invention, and these variations and combinations are still within the protection scope of the present invention.

Claims

1. An echo cancellation processing device for a loud speaker scene, characterized in that: Installed inside the device, including an audio acquisition module, an adaptive threshold compression module, a delay determination module, a delay compensation module, an echo cancellation module and an audio playback module; First, the locally stored audio file is played, and the audio acquisition module collects the echo signal, which is input into the delay determination module, and the delay τ is output and written into the storage medium that is not lost when the power is off; Then, when the device is running online, the audio acquisition module collects the far-end reference signal and the near-end speaker signal sent by the far-end device, and inputs them into the adaptive threshold compression module for compression; at the same time, according to the delay τ stored in advance, the far-end reference signal is delayed compensated by the delay compensation module; finally, the far-end reference signal and the reference signal after the compensation delay are echo-cancelled by the echo cancellation module.

2. An echo cancellation processing method using the echo cancellation processing device according to claim 1, characterized in that: The following steps are involved: Step 1: For the calling device A that is in a call, the audio acquisition module collects the remote reference signal and near-end microphone signal N represents the number of sample points of the signal, which is set according to engineering experience; Step 2: The adaptive threshold compression module adaptively compresses the amplitude of the microphone signal d in the time domain to obtain the signal Adaptive compression includes the following processing steps: Step 201: Calculate the RMS value x of the remote reference signal x rms , and converted to intensity decibel value x dB Right now x=[x0,x1,…,x i ,…,x N-1 ], x[i] represents the i-th element in the remote reference signal x; Step 202: Determine the intensity decibel value x dB Is it greater than the threshold T1? If yes, execute step 203; Otherwise, execute step 206; T1 represents the threshold for determining whether there is a speaker signal at the far end, which is preset based on engineering experience; Step 203, compress the near-end microphone signal d to obtain a compressed near-end microphone signal d′; The specific compression includes the following steps: Step I: according to the strength decibel value d of the near-end microphone signal dB , combined with the preset compression threshold T2 and compression ratio R, the gain g is calculated: The compression threshold T2 and the compression ratio R are preset based on engineering experience; Step II, using the gain g to calculate the compressed near-end microphone collected signal d′, Where d[i] represents the i-th element in the near-end microphone signal d, where 0≤i≤N-1; Step 204: performing a limiting process on the compressed near-end microphone collected signal d′, the specific formula is: Where, 0≤i<N, T3 is the amplitude limiting threshold, and α is the amplitude attenuation coefficient; T3 and α are preset according to engineering experience; Step 205: using the signal d′ collected by the near-end microphone after limiting, the compression threshold T2 is updated as the next compression threshold; Step 206: Do not perform any processing on the signal collected by the near-end microphone, that is, the compressed near-end microphone signal d′ is the near-end microphone signal d; Step 3: Device A runs a delay determination module to obtain a delay τ value of its own device and compensates for a remote reference signal x; Step 4: Perform block-segment frequency domain adaptive filtering based on the signal d′ after compression and limiting processing and the far-end reference signal x after delay compensation.

3. The echo cancellation processing method according to claim 2, characterized in that: In step 205, the specific process of updating the compression threshold T2 is as follows: First, the RMS value d is calculated by using the signal d′ collected by the near-end microphone after limiting. r ' ms , and converted to intensity decibel value d d ' B ; Then, calculate the decibel value x of the remote reference signal dB The intensity decibel value d of the signal collected by the near-end microphone after limiting d ' B The signal strength difference between them is e; As shown below: e = d d ' B -x dB ; Next, when the signal strength difference e is greater than the threshold T4, the near-end microphone is used to collect the signal strength decibel value d after limiting. d ' B Update compression threshold T2: The update formula is as follows: T2 = T2-(d d ' B -x dB )×μ; T4 is the threshold for determining signal strength difference, which is preset based on engineering experience; μ is the update step size, which is preset based on engineering experience.

4. The echo cancellation processing method according to claim 2, characterized in that: In step 3, the delay compensation is performed only on the remote reference signal x containing N sample points in the first frame, specifically: According to the delay τ pre-set in device A and the remote reference signal sampling rate f, the number of delayed sample points is calculated: K = f × τ; Add K zeros to the front end of the remote reference signal x, that is, Take the first N sample points of the far-end reference signal after the front-end is padded with 0, that is, The remaining K sample points are used as the first K sample points of the far-end reference signal containing N sample points in the next frame.

5. The echo cancellation processing method according to claim 2, characterized in that: In step 3, the delay τ of device A is set. The specific steps are as follows: First, in offline state, the speaker of device A plays the local audio file, and the microphone collects the signal And perform adaptive compression processing to obtain the compressed microphone acquisition signal Then, the local audio file is used as a reference signal According to the pre-set candidate delay set {τ1,τ2,…,τ m }, and perform delay compensation on the reference signals respectively to obtain {r1,r2,…,r m }; Where m is the number of candidate delays, delay τ i Corresponding to the compensation signal r i , i∈[1,m]; Next, the compressed signal s′ and the compensated signal r i , respectively, perform frequency domain adaptive filtering in each block segment to obtain the linear filtering result {y1,y2,…,y m }; Finally, according to the linear filtering result {y1,y2,…,y m } and the compressed signal s′, respectively calculate the echo return loss enhancement, and obtain {l1,l2,…,l m }; According to the set {l1,l2,…,l m }, the index i corresponding to the maximum value in the candidate delay set {τ1,τ2,…,τ m }, i.e., τ = τ i As the delay of device A's setup.

Images