CN110444222B - Voice noise reduction method based on information entropy weighting - Google Patents
Voice noise reduction method based on information entropy weighting Download PDFInfo
- Publication number
- CN110444222B CN110444222B CN201910413996.1A CN201910413996A CN110444222B CN 110444222 B CN110444222 B CN 110444222B CN 201910413996 A CN201910413996 A CN 201910413996A CN 110444222 B CN110444222 B CN 110444222B
- Authority
- CN
- China
- Prior art keywords
- voice
- signal
- weighting
- noise reduction
- signals
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0208—Noise filtering
- G10L21/0216—Noise filtering characterised by the method used for estimating noise
Landscapes
- Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Quality & Reliability (AREA)
- Signal Processing (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Noise Elimination (AREA)
Abstract
The invention discloses a voice noise reduction method based on information entropy weighting, which comprises the following steps: step one, demodulating a received signal and then performing framing processing; step two, calculating a power spectrum by using continuous M frames of signals; extracting components in a voice frequency range in the power spectrum and calculating the normalized power spectral density of the components; fourthly, performing spectral entropy estimation on all frequency components in the continuous M frame signals; step five, calculating a voice attribute value; and step six, weighting the voice signals by using the voice attribute values. Compared with the prior art, the invention has the following positive effects: the invention distinguishes whether the received signal is voice or noise by using the characteristic that the spectral entropy of the voice signal and the noise signal has obvious difference, and realizes the amplification of the voice signal and the suppression of the noise by weighting the received signal by using the spectral entropy value.
Description
Technical Field
The invention relates to a voice noise reduction method based on information entropy weighting.
Background
During transmission of signals, noise and interference are important factors affecting the signals. In order to effectively retain the useful signal, remove the background noise and the interference generated in the transmission, and make the receiving party have good listening feeling, it is necessary to perform voice noise reduction processing on the signal at the receiving end. Conventional voice stations typically employ spectral subtraction for noise reduction: and subtracting the estimation value of the noise power spectrum from the voice power spectrum containing the noise, and inversely transforming the residual signal energy back to the time domain to obtain the voice signal after noise reduction. The method is equivalent to that certain equalization processing is carried out on the signal with noise in the frequency domain, amplitude information of specific frequency points is lost, and due to random distribution of noise, the suppression of the noise is not thorough; under the condition of low signal-to-noise ratio, the spectral subtraction method can cause excessive suppression to voice signals, and the subsequent processing and communication quality are affected.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides a voice noise reduction method based on information entropy weighting, which not only provides a calculation method for demodulating the spectral entropy of a voice signal, but also further provides a noise suppression method for spectral entropy weighting.
The technical scheme adopted by the invention is as follows: a voice noise reduction method based on information entropy weighting comprises the following steps:
step one, demodulating a received signal and then performing framing processing;
step two, calculating a power spectrum by using continuous M frames of signals;
extracting components in a voice frequency range in the power spectrum and calculating the normalized power spectral density of the components;
performing spectral entropy estimation on all frequency components in the continuous M frame signals;
step five, calculating a voice attribute value;
and step six, weighting the voice signals by using the voice attribute values.
Compared with the prior art, the invention has the following positive effects:
the invention distinguishes whether the received signal is voice or noise by using the characteristic that the spectral entropy of the voice signal and the noise signal has obvious difference, and realizes the amplification of the voice signal and the suppression of the noise by weighting the received signal by using the spectral entropy value.
Drawings
The invention will now be described, by way of example, with reference to the accompanying drawings, in which:
FIG. 1 is a signal processing flow for speech noise reduction based on information entropy weighting;
FIG. 2 is a simulation result of voice noise reduction using entropy weighting.
Detailed Description
The invention provides a voice noise reduction method based on spectral entropy weighting, wherein the spectral entropy refers to Shannon information entropy of a baseband signal power spectrum of a power station. The voice noise reduction method based on spectral entropy weighting distinguishes whether a received signal is voice or noise by using the characteristic that the size of the spectral entropy of a voice signal is obviously different from that of a noise signal, and realizes the amplification of the voice signal and the suppression of the noise by weighting the received signal by using the value of the spectral entropy.
The received demodulated digital signal x (n), n ═ 1,2, …, uses the way of speech frame-dividing process, the time length of each frame signal is adapted to the time length of one syllable of speech signal, the time length of one syllable is usually between 3ms and 30ms, here, the concrete speech frame length can be selected in this range according to the need. Assuming that the number of samples of one frame signal is N, the k-th frame signal can be expressed as
xk＝[x((k-1)N+1),x((k-1)N+2),…,x(kN)]T,k＝1,2,…
Fourier transform is carried out by utilizing signals of two adjacent frames of the k and the k +1 to obtain
Fourier transform matrix writeable
Wherein N ═ 0,1, 2.. 2N-1]T. The power spectrum of the received signal is calculated using the fourier transform of the successive frame signals as follows:
wherein "o" denotes multiplication of corresponding elements of the vector. Since the voice frequency is rarely higher than 3.5KHz, the frequency component power spectrum lower than 3.5KHz can be selected
To calculate the spectral entropy of the speech signal portion, assuming
The sample length is L. For the influence of the signal power magnitude on the signal spectrum entropy calculation, the spectrum entropy is calculated by using a normalized power spectrum density which can be expressed as
Here, the
Is a column vector of L elements, the L-th element of which is represented as
The spectral entropy of the k frame signal is estimated as
Assuming that the spectral entropy decision threshold is η, the voice attribute variable F of the frame signal can be decided as followsk，
Fk1 means that the k-th frame signal has a speech feature, Fk< 1 indicates that the signal of the k-th frame does not have voice feature, FkAs the weighting coefficient of the output audio signal, the noise suppression capability of the output signal can be improved.
As shown in fig. 1, the method of the present invention comprises the steps of:
1) demodulating the received signal and then performing framing processing;
2) calculating a power spectrum using the successive M frames of signals;
3) extracting the components in the voice frequency range in the power spectrum and calculating the normalized power spectral density of the components;
4) calculating spectral entropy by using the calculated power spectral density, and taking the spectral entropy as a spectral entropy estimation result of a first frame signal in the continuous M frames of signals;
5) calculating a voice attribute value Fk；
6) Weighting the voice signal with the voice attributes.
The following are the implementation examples and simulation results of the method of the present invention:
assuming that an audio signal sampling frequency of 8kHz is adopted, a signal power spectrum is calculated by using a time length of 64ms every 8ms of a frame signal, and voice attribute values of continuous 10 frames of signals are adopted to realize the functions of noise opening and noise closing, and the noise reduction effect is shown in FIG. 2.
Claims (7)
1. A voice noise reduction method based on information entropy weighting is characterized in that: the method comprises the following steps:
step one, demodulating a received signal and then performing framing processing;
step two, calculating a power spectrum by using continuous M frames of signals;
extracting components in a voice frequency range in the power spectrum and calculating the normalized power spectral density of the components;
performing spectral entropy estimation on all frequency components in the continuous M frames of signals;
step five, calculating a voice attribute value:
judging whether the requirements are met
If so, then
Otherwise, Fk1 is ═ 1; wherein,
representing the spectral entropy estimation result of the kth frame signal, wherein eta is a spectral entropy judgment threshold;
and step six, weighting the voice signals by using the voice attribute values.
2. A method of speech noise reduction based on information entropy weighting according to claim 1, wherein: step one the method for performing framing processing is as follows: assuming that the received demodulated digital signal is x (N), N is 1,2, …, selecting the duration of each frame signal in a syllable duration range of the speech signal by adopting a speech framing processing mode, assuming that the number of samples of a frame signal is N, and expressing the k-th frame signal as x (N)k＝[x((k-1)N+1),x((k-1)N+2),…,x(kN)]T,k＝1,2,…。
3. A method of speech noise reduction based on information entropy weighting according to claim 2, wherein: the duration of the one syllable is 3ms to 30 ms.
4. A method of speech noise reduction based on information entropy weighting according to claim 2, wherein: step two the method for calculating the power spectrum by using the continuous M frame signals comprises the following steps:
1) fourier transform is carried out by utilizing signals of two adjacent frames of the k and the k +1 to obtain
The Fourier transform matrix F in the formula is calculated according to the following formula:
wherein N ═ 0,1, 2.. 2N-1]T；
2) The power spectrum of the received signal is calculated using the fourier transform of the successive frame signals as follows:
therein
Representing the multiplication of corresponding elements of the vector.
5. A method of speech noise reduction based on information entropy weighting according to claim 4, wherein: step three, the method for calculating the normalized power spectral density comprises the following steps: selecting a frequency component power spectrum with voice frequencies below 3.5KHz
Suppose that
The length of the sample is L, and the normalized power spectral density is calculated according to the following formula:
therein
Is a column vector of L elements, the L-th element of which is
6. A method of speech noise reduction based on information entropy weighting according to claim 5, wherein: the method for estimating the spectral entropy comprises the following steps: will be provided with
The L-th element of
According to the following formulaCalculating to obtain a spectral entropy estimation result of the kth frame signal:
7. a method of speech noise reduction based on information entropy weighting according to claim 1, wherein: step six the method for weighting the voice signal by using the voice attribute value comprises the following steps: when F is presentkWhen 1, the k frame signal has a voice feature; when F is presentk<When 1, the signal of the k frame does not have the voice feature, FkThe noise suppression is implemented as a weighting factor of the output audio signal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910413996.1A CN110444222B (en) | 2019-05-17 | 2019-05-17 | Voice noise reduction method based on information entropy weighting |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910413996.1A CN110444222B (en) | 2019-05-17 | 2019-05-17 | Voice noise reduction method based on information entropy weighting |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110444222A CN110444222A (en) | 2019-11-12 |
| CN110444222B true CN110444222B (en) | 2021-12-14 |
Family
ID=68429076
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910413996.1A Active CN110444222B (en) | 2019-05-17 | 2019-05-17 | Voice noise reduction method based on information entropy weighting |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110444222B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11270720B2 (en) * | 2019-12-30 | 2022-03-08 | Texas Instruments Incorporated | Background noise estimation and voice activity detection system |
| CN114885252A (en) * | 2022-03-30 | 2022-08-09 | 广州市迪声音响有限公司 | Volume detection method and detection system for digital sound console |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103426440A (en) * | 2013-08-22 | 2013-12-04 | 厦门大学 | Voice endpoint detection device and voice endpoint detection method utilizing energy spectrum entropy spatial information |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1459761B (en) * | 2002-05-24 | 2010-04-21 | 清华大学 | Character Recognition Technology Based on Gabor Filter Bank |
| CN101599269B (en) * | 2009-07-02 | 2011-07-20 | 中国农业大学 | Phonetic end point detection method and device therefor |
| CN103945214B (en) * | 2013-01-23 | 2016-03-30 | 中兴通讯股份有限公司 | End side time-domain method for evaluating video quality and device |
| CN107257528A (en) * | 2017-06-14 | 2017-10-17 | 山东浪潮云服务信息科技有限公司 | A kind of detection method of uttering long and high-pitched sounds based on weighted spectral entropy |
-
2019
- 2019-05-17 CN CN201910413996.1A patent/CN110444222B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103426440A (en) * | 2013-08-22 | 2013-12-04 | 厦门大学 | Voice endpoint detection device and voice endpoint detection method utilizing energy spectrum entropy spatial information |
Non-Patent Citations (1)
| Title |
|---|
| 一种改进的自适应子带谱熵语音端点检测方法;刘华平等;《系统仿真学报》;20080305(第05期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110444222A (en) | 2019-11-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108831499B (en) | Speech enhancement method using speech existence probability | |
| US8010355B2 (en) | Low complexity noise reduction method | |
| RU2127454C1 (en) | Method for noise suppression | |
| CN110634500B (en) | Method for calculating prior signal-to-noise ratio, electronic device and storage medium | |
| CN101853666B (en) | Speech enhancement method and device | |
| CN104067339B (en) | Noise-suppressing device | |
| WO2022160593A1 (en) | Speech enhancement method, apparatus and system, and computer-readable storage medium | |
| JP2002541753A (en) | Signal Noise Reduction by Time Domain Spectral Subtraction Using Fixed Filter | |
| JP2003534570A (en) | How to suppress noise in adaptive beamformers | |
| CN101719969A (en) | Method and system for judging double-end conversation and method and system for eliminating echo | |
| CN107316648A (en) | A kind of sound enhancement method based on coloured noise | |
| CN103109320A (en) | Noise suppression device | |
| CN110444222B (en) | Voice noise reduction method based on information entropy weighting | |
| Yang et al. | A noise reduction method based on LMS adaptive filter of audio signals | |
| CN112634926B (en) | Short wave channel voice anti-fading auxiliary enhancement method based on convolutional neural network | |
| CN113593599A (en) | Method for removing noise signal in voice signal | |
| CN106340302A (en) | De-reverberation method and device for speech data | |
| CN112151060B (en) | Single-channel voice enhancement method and device, storage medium and terminal | |
| JP2004341339A (en) | Noise restriction device | |
| CN109087657B (en) | Voice enhancement method applied to ultra-short wave radio station | |
| CN107045874B (en) | Non-linear voice enhancement method based on correlation | |
| CN112634927B (en) | Short wave channel voice enhancement method | |
| CN114401168B (en) | Voice enhancement method applicable to short wave Morse signal under complex strong noise environment | |
| CN111933169B (en) | Voice noise reduction method for secondarily utilizing voice existence probability | |
| US7177805B1 (en) | Simplified noise suppression circuit |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |