WO2006017993A1

WO2006017993A1 - A background noise eliminate device and method for speech communication terminal

Info

Publication number: WO2006017993A1
Application number: PCT/CN2005/001289
Authority: WO
Inventors: Qiang Zhang
Original assignee: Huawei Technologies Co., Ltd.
Priority date: 2004-08-18
Filing date: 2005-08-18
Publication date: 2006-02-23
Also published as: CN1737905A; EP1768109A4; CN100337270C; EP1768109A1

Abstract

A background noise eliminate device for speech communication terminal, is consisted of a sound sensor, an adaptive filter, a noise estimator and an adder connected successively. A background noise eliminate method for speech communication terminal includes the following steps: a. collecting the sound signal listened by the receiver which includes the speech of the transmitter and the background noise at the receiver; b. estimating the background noise signal at the receiver from the signal acquired from the step a; c. adding the estimated background noise signal at the receiver after the phase of it being reversed to the speech signal of the transmitter received by the speech communication terminal, and playing the added signal to the receiver.

Description

Device and method for eliminating background noise of voice communication terminal

Technical field

The present invention relates to a communication terminal voice processing technology, and in particular to a device and method for eliminating background noise of a voice communication terminal. Background of the invention

With the rapid development and increasing popularity of communication technologies, communication has become an important means of daily communication and communication. Therefore, people have put forward higher requirements for the voice quality of calls. In actual communication, both communication users often talk in some noisy background environment. At this time, the sound heard by the recipient user includes not only the voice of the sender user, but also other voices around the sender and the receiver, such as voices of others, footsteps, collision of objects, music, and sounds from vehicles. . These sounds other than the speech voices of the communication users are called background noise, the background noise around the sender's user is called the sender's background noise, and the background noise around the receiver's user is called the receiver's background noise.

1 is a schematic diagram of a voice transmission process of a mobile communication terminal, wherein a sender microphone simultaneously receives a sender voice and a sender background noise, and after transmitting, encoding, transmitting, and decoding the received voice, transmitting to the receiver Headphones or speakers, the receiver listens to the transmitted sound through headphones or speakers, and also hears the receiver's background noise. It can be seen that the speech receives double interference of the sender's background noise and the receiver's background noise during the transmission process. In many cases, the background noise masks the voice of the communicating speaker, making the recipient's listener unable to correctly acquire the voice information sent by the speaker, thereby causing communication difficulties.

Based on the voice delivery process of the mobile communication terminal shown in FIG. 1, existing mobile communication terminals, such as CDMA mobile phones and GSM mobile phones, before the vocoder of the utterer encodes the voice, Adaptive filtering techniques are used to eliminate the sender's background noise. 2 is a schematic structural diagram of a device for canceling a background noise of a sender in a mobile communication terminal of the prior art, including: a microphone 201, a sampling module 202, an adaptive filter 203, and a speech encoding module 204.

The microphone 201 receives the sender voice and the background noise, and then transmits the received sender voice and background noise to the sampling module 202. After receiving the voice and background noise, the sample module 202 samples the sample, which is to be simulated. The signal is converted into a digital signal, and the obtained digital signal is output to the adaptive filter 203; the adaptive filter 203 performs background noise cancellation on the received signal, and outputs the remaining speech signal to the speech encoding module 204; Speech encoding module 204 encodes the received speech signal.

The specific structure of the adaptive filter 203 is shown in FIG. 3. Where x(n) is the input signal received by the sampled microphone, including the sender voice signal s(n) and the sender background noise signal v(n), ie x(n) = s(n) + v (n). After x(n) is delayed by Δ, χ(η-Δ)= s(nA) + v(nA) is obtained, and χ(η-Δ) passes through the filter portion of the adaptive filter 203, and the speech signal s(n) is output. The estimated value, and the estimated value (/7) is subtracted from the input signal x(n) to obtain an error signal e(n). The subtraction operation is not a general algebraic subtraction, but requires a corresponding algorithm to perform operations, such as power spectral density analysis of the relevant power.

The adaptive filter 203 employs a Least Mean Square (LMS) algorithm whose weight is determined by the error signal e(n) = x(n)-;(") = v(n) + (s(n) -;(") ) adjustment, where 0 ≤ ≤ M - 1 , M is the order of the filter. If the weight vector W(n) and the input vector X n) are defined as:

f (") = [w.(") w _x (n) ■■■ w _M _ ₂ (n) w _M _ _x (n)

X(n) = [x ₀ (n) χ (η) · · · x _M _ ₂ (n)

Then the LMS algorithm of the adaptive filter can be expressed as:

s(n) = W ^T (n)X(n - Δ) = W ^T (n)S(n - Δ) + W ^T (n)V(n - Δ) e(n)=x(n) - s(n)

W(n+1) = W(n) + μΧ(η)θ(η) ₀

Where μ is the step factor when searching according to the minimum mean square error criterion.

Since the speech has quasi-periodicity, the speech signals s(n) and s(nA) are strongly correlated, and the noise signals v(n) and ν(η-Δ) are considered irrelevant, and the noise signal and the speech signal are also considered uncorrelated. . Based on this assumption, a highly correlated component of the input signal χ(η) can be estimated. At the same time, adjusting the weight of the filter » to minimize the mean square error of e(n), the best estimate of the speech signal s(n) under the minimum mean square error criterion can be obtained.

Although the prior art scheme for eliminating the background noise of the sender can reduce the interference of the background noise to the voice, the prior art only considers to eliminate the background noise of the sender, and does not perform any processing on the background noise of the receiver, ignoring the receiver. The background noise interferes with the sound emitted by the receiver's earphones, especially when the receiver is in a noisy environment, and the receiver's background noise is particularly severe to the speaker's voice. At present, the method for the mobile communication terminal to overcome the background noise of the receiver is to increase the volume of the earphone. This method is limited by the maximum volume of the earphone, and the effect in the noisy environment is still poor. Similarly, the existing fixed communication terminal also has the above problems. Summary of the invention

In view of this, the main object of the present invention is to provide a device for canceling the background noise of a voice communication terminal, which can perform noise cancellation processing on the background noise of the receiver and reduce interference of the receiver background noise on the receiver.

Another object of the present invention is to provide a method for canceling background noise of a voice communication terminal, which can reduce interference of receiver background noise to a receiver.

Based on the above first main object, the present invention provides a device for canceling background noise of a voice communication terminal, the voice communication terminal including for converting an input sound signal into a player that plays sounds;

The device includes:

a sound sensing module, configured to collect sound played by the player and background noise of the receiver, convert the sound into a sound signal, and output the sound signal;

An adaptive filter, configured to calculate and output a noise signal contained in a sound signal of the sound sensing module according to the input sound signal from the sound sensing module and the voice signal from the transmitting side;

a noise estimation module, configured to predict and output a receiver background noise signal according to the noise signal output by the adaptive filter;

And an adder module, configured to inversely superimpose the receiver background noise signal output by the noise estimation module and the sender voice signal into the player of the voice communication terminal.

The signal output by the sound sensing module of the device is an analog signal; the device further includes: an A/D conversion module, configured to convert an analog signal output by the sound sensing module into a digital signal and input the Filter

The number of the sound sensing modules of the device is two or more.

The device further includes an inserter;

The adder is located between the sound sensing module and the A/D conversion module, and is configured to superimpose and output the analog signals output by the plurality of sound sensing modules to the A/D conversion module;

Or the adder is located between the A/D conversion module and the adaptive filter, and is used for outputting the plurality of sound sensing modules to the A/D conversion module and performing analog-to-digital conversion by the A/D conversion module respectively. The signals are superimposed and output to an adaptive filter.

The A/D conversion module of the device is an A/D conversion module for respectively performing analog-to-digital conversion on signals output by a plurality of sound sensing modules and outputting the signals to an adaptive filter respectively; Receiving multiple signals output by the A/D conversion module and filtering separately Wave adaptive filter.

Based on the above another main object, the present invention provides a method for eliminating background noise of a voice communication terminal, including:

a. collecting a sound signal that is received by the recipient user and includes the sender's voice and the receiver's background noise;

b. estimating the receiver background noise signal from the signal obtained in step a;

c The estimated background noise signal of the receiver is reversed and then superimposed with the sender voice signal received by the voice communication terminal, and then played to the receiver user.

The method for collecting the sound signal in the method step a is: converting the collected sound into an analog signal and converting it into a digital signal through analog-to-digital conversion.

The sound collected in the method step a is divided into multiple channels, and after converting the collected sound into an analog signal, the method further includes:

Performing analog-to-digital conversion by superimposing multiple analog signals;

Or respectively, multi-channel analog signals are subjected to analog-to-digital conversion, and the conversion results are superimposed; or the multi-channel analog signals are respectively subjected to analog-to-digital conversion, and respectively output.

The method step b includes: using the sender voice signal received by the voice communication terminal as a reference signal for adaptive filtering processing, using the adaptive filtering processing calculation step a to obtain the noise signal estimation value in the signal, and then using the noise signal estimation value The estimated value of the receiver's background noise signal is predicted.

The method for obtaining a noise signal by using adaptive filtering processing in the method is: obtaining an estimated value of the noise signal by using a minimum mean square error algorithm.

The signal obtained in step a of the method is multiplexed, and the multiple input signals are respectively adaptively filtered in step b, and the maximum value of the plurality of processing results obtained through the adaptive filtering process or all the processing results of the current time are selected. The sum is used as an estimate of the noise signal.

The method for predicting a receiver background noise signal estimate by the method is: current time The background noise signal estimate is the sum of the background noise signal estimate at the previous moment and the noise signal estimate obtained through the adaptive filtering process.

It can be seen from the above scheme that the key to the present invention is that the apparatus of the present invention mainly includes an adaptive filter and a noise estimation module for estimating background noise, and the background noise is estimated by the adaptive filter and the noise estimation module, and the background noise is used. After the estimated value is inverted, the voice signal is superimposed, and then the voice signal including the background noise estimate is played out through the earphone, so that the background noise estimate and the actual receiver background noise cancel each other, thereby obtaining the noise cancellation. Voice signal.

Therefore, the apparatus and method for eliminating background noise of the voice communication terminal provided by the present invention can well cancel the background noise around the earphone, reduce the interference of the receiver background noise to the receiver, and make the receiving party have a small volume in the earphone. In the case of the sender, the voice of the sender can be clearly heard. BRIEF DESCRIPTION OF THE DRAWINGS

1 is a schematic diagram of a voice transmission process of a mobile communication terminal;

2 is a schematic structural diagram of a structure of a device for canceling a background noise of a sender in a mobile communication terminal of the prior art;

3 is a schematic structural diagram of an adaptive filter in a transmitter background noise canceling apparatus shown in FIG. 2;

4 is a schematic structural diagram of an embodiment of a device for canceling a background noise of a receiver in a mobile communication terminal according to the present invention;

FIG. 5 is a schematic structural diagram of an adaptive filter in an embodiment of the receiver background noise canceling apparatus shown in FIG. 4; FIG.

6 is a schematic diagram showing an arrangement manner of an earphone and a microphone in an embodiment of a receiving device background noise canceling apparatus according to the present invention; 7 is a flowchart of a method for implementing background noise for eliminating a voice communication terminal according to the present invention;

Figure 8 is a block diagram showing the structure of an embodiment of a device for canceling background noise of a receiving side in a fixed communication terminal of the present invention. Mode for carrying out the invention

The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

The voice communication terminal to which the present invention is applied should include a player for converting the input sound signal into sound for playback. The device for eliminating background noise of a voice communication terminal provided by the invention mainly comprises: a sound sensing module, configured to collect sound played by a player of a voice communication terminal and background noise of a receiver, convert the collected sound into a sound signal and output; a filter, configured to calculate, according to the input sound signal from the sound sensing module and the voice signal from the sender, a noise signal contained in the sound signal of the sound sensing module, and output; a noise estimation module, configured to The noise signal output by the adaptive filter predicts the receiver background noise signal and outputs; the adder module is configured to inversely superimpose the receiver background noise signal output by the noise estimation module and the sender voice signal into the a player of the voice communication terminal.

Based on the same principle, the method for eliminating background noise of a voice communication terminal provided by the present invention mainly includes the following steps: collecting a sound signal including a sender voice and a receiver background noise received by a receiver user; and estimating a receiver background from the obtained signal. The noise signal is superimposed on the estimated background noise signal of the receiver and superimposed with the sender voice signal received by the voice communication terminal, and then played to the receiver user.

Referring to FIG. 4, the structure of the device for canceling the background noise of the receiver in the mobile communication terminal of the preferred embodiment of the present invention includes: an adder module 401, a digital/analog (D/A) conversion module 402, an earphone 403, a microphone 405, Analog/digital (A/D) conversion module 406, adaptive filter 407 And a noise estimation module 408, wherein the earphone 403 and the digital/analog (D/A) conversion module 402 are existing devices in the mobile communication terminal.

The adder module 401 is configured to inversely superimpose the sender voice signal s(n) received by the communication terminal and the receiver background noise signal estimate value (7); the D/A conversion module 402 is configured to input the digital signal. Converting to an analog signal; the earphone 403 is configured to convert the input analog electrical signal into a sound signal for playing. In the present invention, the earphone 403 actually plays the voice signal from the sender and the opposite phase of the received background noise signal. After the signal, the speaker can also be used instead of the earphone; the microphone 405 is used for receiving the user's voice signal and the background noise cancelled by the headphone output signal; the A/D conversion module 406 is used to convert the analog signal into a digital signal; Filter 407 is used to estimate the noise signal value in the acquired signal; noise estimation module 408 is used to predict the estimated value of the background noise.

The sender voice signal s(n) received by the communication terminal is processed by the adder module 401 and then transmitted to the D/A conversion module 402, and the D/A conversion module 402 performs digital signal to analog signal conversion on the received signal, and The converted analog signal is output to the earphone 403 as a player, and the earphone 403 converts the analog signal into a sound signal for playback. The sound signal played by the earphone 403 is transmitted to the receiver user and the microphone 405 as the sound sensing module through the voice transmission channel 404, and the background noise V of the receiver is also transmitted to the receiver user and the microphone 405 through the voice transmission channel 404. .

The microphone 405 converts the received sound signal into an analog signal X and sends it to the A/D conversion module 406. The signal X includes the sender's voice signal and the receiver noise signal. The A/D conversion module 406 samples the analog signal and The converted value signal x(n) is sent to the adaptive filter 407; the adaptive filter 407 performs adaptive filtering with the speech signal s(n) as a reference input, and the estimated value noise estimation module 408 of the estimated noise signal is based on the noise signal. The estimated value predicts the estimated value of the background noise signal and sends the background noise estimate The adder module 401 is provided. Here, the A/D conversion module 406 is not necessary, for example: if the adaptive filter 407, the noise estimation module 408, and the like can process the analog signal, the A/D conversion module 406 does not have to be set; in addition, if the microphone 405 can The digital signal is directly output, and it is not necessary to set the A/D conversion module 406.

The adder module 401 superimposes the estimated value of the received background noise signal with the transmitted speech signal s(n) to obtain a signal s(n) - ν(π), and outputs the signal to the D/A. The conversion module 402 is played back through the earphone 403. The inverse phase of the received background noise signal estimate - θ) and the receiver background noise V cancel each other in the voice transmission channel 404 such that the sound transmitted to the recipient user and the microphone 405 is primarily the sender's voice signal.

The specific structure of the adaptive filter 407 is shown in FIG. The mixed digital signal x(n) received via the microphone including the sender speech signal s(n) and the background noise signal passes through the filter portion of the adaptive filter 407 to recover the best estimate of the transmitted speech signal) The best estimated value of the speech signal is subtracted from the delayed signal s(nA) of the transmitted speech signal s(n) received by the mobile terminal through the antenna to obtain the obtained error signal.

The adaptive filter 407 employs an LMS algorithm whose weight ¹ ^") is adjusted by the error signal = s(n) - s(nA), where 0 ≤ ≤ ikT - 1 , M is the order of the filter. The signal is not correlated with the background noise signal. Adjusting the weight of the filter ^w » to minimize the mean square error, the best estimate of the speech signal s(n) in χ(η) under the minimum mean square error criterion can be obtained. At the same time, it is considered that W is the best estimate of the residual noise signal after the background noise is cancelled. Where Δ is the delay of the speech signal, which can be searched according to the minimum mean square error criterion. Of course, the adaptive filter 407 can also Use other adaptive algorithms and best practices.

The noise estimation module 408 predicts the current background based on the best estimate of the noise signal Noise estimate. The simplest prediction method is v(") = v(" _ l) + v _e (") where v(«) is the current background noise estimate and _ 1) is the background noise estimate from the previous moment. .

In order to make the effect of noise cancellation better, the number of headphones 403 and microphone 405 can be set to be plural, and the number of headphones and microphones can be different. The position of the microphone is in the area where it is desired to cancel the noise. The area where it is usually desired to cancel the noise is around the earphone on the receiving side, but there are exceptions in other applications, such as: When sleeping, if the mobile phone is only used as a tool to reduce the surrounding noise, This area can be larger. Taking one earphone and three microphones as an example, the distribution method shown in Fig. 6 can be adopted. Among them, the earphone is in the middle, and the three microphones are distributed at an angle of 120 degrees around the earphone.

Signals received simultaneously by multiple microphones may be superimposed and transmitted to the adaptive filter, or separately transmitted to the adaptive filter and then processed by the adaptive filter separately. If superimposition is required, the corresponding position in the device of the present invention is required. Add a stacker. For example: the output signals of multiple microphones can be superimposed before A/D conversion, then the adder is set between the microphone and the A/D conversion module; or it can be superimposed after A/D conversion. The adder is disposed between the A/D conversion module and the adaptive filter. In addition, the output signals of the plurality of microphones can also be separately sent to the adaptive filter for processing after performing A/D conversion.

The present invention also provides a background noise canceling method for a receiving party voice communication terminal, as shown in Fig. 7, comprising the following steps:

Step 701: Collecting a sound to be heard by the user of the receiving party, including a sound emitted by the earphone of the voice communication terminal and a background noise of the receiving side, and converting the collected sound into a sound signal and performing A/D conversion.

The microphones for collecting sound signals may be one or more, that is, the collected signals may be one or more channels. Multiple microphones can be used when acquiring signals using multiple microphones The signals of the wind output are first superimposed, and then A/D is converted and outputted. The signals output by the plurality of microphones may be respectively A/D-converted and then superimposed and outputted; or the signals output by the plurality of microphones may be separately Perform A/D conversion and output separately.

Here, the A/D conversion process is also not necessary, for example: if the subsequent adaptive filtering, noise estimation, etc. processing can be performed based on the analog signal, the A/D conversion is not necessary here; in addition, if the microphone can directly output the digital The signal does not have to be subjected to the A/D conversion.

Step 702: The sender voice signal is used as a reference input, and the collected digital sound signal is adaptively filtered to estimate a noise signal therein. wherein the adaptive filter may adopt an LMS algorithm or adopt another adaptive algorithm and most Good guidelines.

When the signals output from a plurality of microphones are respectively A/D-converted and output to the adaptive filter, it is necessary to perform adaptive filtering processing on each of the signals. Here is a brief description of three microphones and one earphone.

For example, the signals received by the three microphones are processed by the adaptive filter under the minimum mean square error criterion to obtain the estimated values of the three noise signals ^φ ₇ ), ^ ₂ («) and ^3 (/7). Then calculate v _e 0) =max( (77) , v _e2 (η) , v _e3 (77) ) or v _e {n) =sum( v _e i (n) , v _e2 (;7) , v ( n) ) , where max is the maximum operation and sum is the sum operation. That is to say, the maximum of the three estimated values obtained after the adaptive filtering process can be selected or the three estimated values can be superimposed as the adaptive filtering result output. Since the delays of the signals are different, it is necessary to first evaluate the noise signals of each channel, such as («), ₂ ( ) and _e3 ( ), and then select or superimpose them.

Since the human ear's perception of sound is not proportional to the power of the sound, but proportional to the number of decibels of the sound, it is also possible to take the logarithm of the absolute value of ^("), that is, log to minimize the log. Optimal criteria. Correspondingly, for multiple inputs, The absolute value of each _Va is taken as a logarithm, and then summed, that is, (log| ve(n)), and then ∑(log ; ν _β («)) is minimized as the optimal criterion. Step 703: Predict an estimated value of the actual background noise signal according to the obtained noise signal; the prediction method of the most single unit is: o) = (-1)+ve(n). Where, where is the current estimated value of the background noise signal, /7-l) is the estimated value of the background noise signal at the previous moment, and 7) is the best estimate of the noise signal obtained by adaptive filtering.

Step 704: The predicted background actual signal noise signal is inversely phase-phased and superimposed with the sender voice signal, and output through the earphone.

Step 705: The sound outputted by the earphone and the background noise of the receiver are superimposed during the voice transmission to cancel the background noise therein, and the receiver user obtains the noise of the sender after the noise cancellation.

The solution of the invention can also be applied to a fixed voice communication terminal, and the implementation thereof is only different in that: since the existing fixed communication terminal receiver receives the analog signal from the sender, it is necessary to first convert the received analog signal. A digital signal is then used for noise estimation and noise cancellation. Referring to Figure 8, it is necessary to add an A/D conversion module 800 before the sender's voice signal input terminal for converting the input analog signal into a digital signal.

It can be seen from the above scheme that the present invention can obtain residual noise after noise cancellation through adaptive filtering processing, and then continuously superimposes by noise prediction processing, and can estimate the background noise of the receiver, and then use the estimated receiver background. The noise is cancelled by the actual receiver background noise, which reduces the interference of the background noise to the receiver.

In conclusion, the above description is only the preferred embodiment of the present invention, and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., which are included in the spirit and principle of the present invention, should be included in the present invention. Within the scope of protection of the invention.

Claims

Claim

A device for canceling background noise of a voice communication terminal, the voice communication terminal comprising a player for converting an input sound signal into a sound for playing;

Characterized in that the device comprises:

a sound sensing module, configured to collect sound played by the player and background noise of the receiver, convert the sound signal into a sound signal, and output the sound signal;

2. The device according to claim 1, wherein the signal output by the sound sensing module is an analog signal; the device further comprising: an A/D conversion module, configured to output the sound sensing module The analog signal is converted into a digital signal and input to the adaptive filter

The apparatus according to claim 1, wherein the number of the sound sensing modules is two or more.

4. Apparatus according to claim 3, wherein the apparatus further comprises a stacker;

The adder is located between the sound sensing module and the A/D conversion module, and is configured to superimpose and output the analog signals output by the plurality of sound sensing modules to the A/D conversion module; Or the adder is located between the A/D conversion module and the adaptive filter, and is used for outputting the plurality of sound sensing modules to the A/D conversion module and performing analog-to-digital conversion by the A/D conversion module respectively. The signals are superimposed and output to an adaptive filter.

The device according to claim 3, wherein the A/D conversion module is configured to perform analog-to-digital conversion on signals output by the plurality of sound sensing modules and output the signals to the adaptive filter respectively. The A/D conversion module; the adaptive filter is an adaptive filter for receiving multiple signals output by the A/D conversion module and separately performing filtering processing.

6. A method for canceling background noise of a voice communication terminal, comprising: a. collecting a sound signal that includes a sender voice and a receiver background noise received by a receiver user;

7. The method according to claim 6, wherein the method for collecting the sound signal in the step a is: converting the collected sound into an analog signal and converting it into a digital signal through analog-to-digital conversion.

8. The method according to claim 7, wherein the sound collected in step a is a multi-channel acquisition, and after converting the collected sound into an analog signal, the method further comprises: multi-channel simulation The signal is superimposed and then analog-to-digital converted;

Alternatively, the analog signals of the multiple analog signals are separately converted, and the conversion results are superimposed; or the analog signals of the multiple analog signals are separately converted and output separately.

The method according to any one of claims 6 to 8, wherein the step b comprises: using the sender voice signal received by the voice communication terminal as a reference signal for adaptive filtering processing, and calculating by using adaptive filtering processing Step a obtains an estimate of the noise signal in the signal, The estimate of the received background noise signal is then predicted using the noise signal estimate.

The method according to claim 9, wherein the method for obtaining a noise signal by using adaptive filtering is: obtaining an estimated value of the noise signal by using a minimum mean square error algorithm. '

The method according to claim 9, wherein the signal obtained in step a is multiplexed, and the multiple input signals are separately adaptively filtered in step b, and the obtained adaptive filtering process is selected. The maximum value of the processing results or the sum of all the processing results of this time is used as an estimated value of the noise signal.

12. The method according to claim 9, wherein the method for predicting the estimated value of the background noise signal of the receiver is: the estimated value of the background noise signal at the current time is the estimated value of the background noise signal at the previous moment and The sum of the noise signal estimates obtained by adaptive filtering.