CN105304087B - Voiceprint recognition method based on zero-crossing separating points - Google Patents
Voiceprint recognition method based on zero-crossing separating points Download PDFInfo
- Publication number
- CN105304087B CN105304087B CN201510586504.0A CN201510586504A CN105304087B CN 105304087 B CN105304087 B CN 105304087B CN 201510586504 A CN201510586504 A CN 201510586504A CN 105304087 B CN105304087 B CN 105304087B
- Authority
- CN
- China
- Prior art keywords
- zero
- matrix
- characteristic vector
- zero crossing
- statistics
- 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
Landscapes
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
- Image Analysis (AREA)
- Collating Specific Patterns (AREA)
Abstract
The invention discloses a voiceprint recognition method. The method has the advantages of simple process, small calculation quantity and high precision of the voiceprint recognition, and has the concrete processes of collecting a sound signal and determining zero-crossing points; counting the number of sampling points between all adjacent zero-crossing points to build a one-dimension feature vector; counting the number of the sampling points between all of the zero-crossing points separated by one zero-crossing point to build a two-dimensional feature vector; deducing the rest points by analogy to obtain a multi-dimension feature vector of the sound signal; building a template base; and realizing the voiceprint recognition through the matching of the multi-dimensional feature vector of the detected sound signal and the template base.
Description
Technical field
The invention belongs to computer and information services field, and in particular to a kind of vocal print based on zero passage spaced points is known
Other method.
Background technology
Sound groove recognition technology in e be 20th century mid-term the U.S. propose, carry out technique research earliest is U.S. shellfish
The Lao Lunsikesite of your laboratory, he has carried out Analysis and Identification by the up to ten thousand vocal print figures to more than 100 Healthy Peoples, accurately
Rate reaches 99.65%.The sound groove recognition technology in e of China is started late, and just proceeds by formal research in the nineties in 20th century,
Carry out correlational study at present has Peking University, Tsing-Hua University, Acoustical Inst., Chinese Academy of Sciences and some political-legal departments.With
The continuous development of each research field (such as material, communication, computer, life sciences etc.), sound groove recognition technology in e is also at full speed
Development, its reliability and accuracy will be improved constantly.Mainly include with regard to the algorithm of Application on Voiceprint Recognition：Mel cepstrum coefficients (MFCC)
Method, based on the method for HMM (HMM), method based on zero passage spaced points etc..
MFCC and HMM methods accuracy of identification is high, but their amount of calculation is too big, high to hardware requirement.Based between zero passage
The method amount of calculation of dot interlace is little, it is only necessary to which relatively low sample rate and less sampled point can just complete the identification function of vocal print,
But accuracy of identification is low.
The content of the invention
In view of this, the invention provides a kind of method for recognizing sound-groove based on zero passage spaced points, the method will be original
One-dimensional zero passage spaced points recognition methodss expand to multidimensional identification, improve accuracy of identification.
Realize that technical scheme is as follows：
A kind of method for recognizing sound-groove based on zero passage spaced points, detailed process is：
S00, collected sound signal：
Sample rate is set as n, one section of acoustical signal is sampled, sampled point number is k；
S01, determine zero crossing：
If the sampled value when not having acoustical signal is X, when having acoustical signal, the value of each sampled point is X (1), X (2) ...
X (k), when formula (1) is met：
(X(i)-X)(X(i+1)-X)≤0(i≤k-1) (1)
Then remember that X (i) is zero crossing, by that analogy, records all zero crossings；
The sampled value of all zero crossings is designated as into y (1), y (2) ... y (ε), wherein ε is the sum of all zero crossings；
S02, statistics zero passage spaced points：
First, the number of the sampled point between all adjacent zero crossings is counted, that is, is counted y (i+1) and is sampled and y (i) between
The number of point, and be stored in matrix z1, wherein i=1,2 ... ε -1；The number of times that each element occurs in statistical matrix z1,
And store the result of statistics in matrix w1, using w1 as the first dimensional feature vector；
Secondly, the number of the sampled point between the zero crossing of one zero crossing in all intervals is counted, that is, counts y (i+2) and y
The number of sampled point between (i), and be stored in matrix z2, wherein i=1,2 ... ε -2；Each element in statistical matrix z2
The number of times of appearance, and the result of statistics is stored in matrix w2, using w2 as the second dimensional feature vector；
By that analogy, obtain successively be separated by two, three ..., sampled point between the zero crossing of N-1 zero crossing
Number, obtain w3, w4 ..., wN；
S03, set up multidimensional characteristic matrix：
By w1, w2 ..., the short characteristic vector subsequent Zero of length in wN, obtain the characteristic vector of a N-dimensional；
S04, set up template base：
According to the mode of step S00 to S03, its corresponding N-dimensional feature is obtained respectively for various different acoustical signals
Vector, builds template base；
S05, obtain matching result：
According to the mode of step S00 to S03, the N-dimensional characteristic vector for being detected acoustical signal is extracted, and by itself and template base
The characteristic vector of middle acoustical signal is matched, and realizes the identification of vocal print.
Further, the process that matched of the present invention is：By the N-dimensional characteristic vector of detected acoustical signal and mould
In plate storehouse, characteristic vector asks for Euclidean distance respectively, if the Euclidean distance of minimum is less than the threshold value of setting, by minimum euclidean distance
The signal being detected otherwise, is considered as unknown signaling as the acoustical signal for matching by corresponding acoustical signal.
Beneficial effect：
Method provided by the present invention obtain successively be separated by one, two, three ..., the zero crossing of N-1 zero crossing it
Between sampled point number, set up multidimensional characteristic matrix to be matched, with Mel cepstrum coefficients (MFCC) method, based on hidden Ma Er
The method of section's husband's model (HMM) compares that amount of calculation is little, and compared with traditional voiceprint recognition algorithm, real-time is good, high precision.
Description of the drawings
Fig. 1 is the flow chart of method provided by the present invention.
Specific embodiment
Below in conjunction with the accompanying drawings, describe the present invention.
The invention provides a kind of method of Application on Voiceprint Recognition, as shown in figure 1, the method is concretely comprised the following steps：
S00, collected sound signal：
Sample rate is set as 3000HZ, one section of acoustical signal is sampled, sampled point number is 200.
S01, determine zero crossing：
If the sampled value when not having acoustical signal is X, when having acoustical signal, the value of each sampled point is X (1), X (2) ...
X(k).When formula below is met：
(X(i)-X)(X(i+1)-X)≤0(i≤k-1) (1)
Then remember that X (i) is zero crossing, by that analogy, records all zero crossings.
The sampled value of all zero crossings is designated as into y (1), y (2) ... y (ε), wherein ε is the sum of all zero crossings.
S02, statistics zero passage spaced points：
First, the number of the sampled point between all adjacent zero crossings is counted, that is, is counted y (i+1) and is sampled and y (i) between
The number of point, and be stored in matrix z1, wherein i=1,2 ... ε -1；The number of times that each element occurs in statistical matrix z1,
And store the result of statistics in matrix w1, using w1 as the first dimensional feature vector；
For example, the element for occurring in matrix z1 is respectively α1、α2……αk；α is counted respectively1、α2……αkOccur in z1
Number of times, be designated as w1 (α1)、w1(α2)……w1(αk), then using w1 as the first dimensional feature vector.
Secondly, the number of the sampled point between the zero crossing of one zero crossing in all intervals is counted, that is, counts y (i+2) and y
The number of sampled point between (i), and be stored in matrix z2, wherein i=1,2 ... ε -2；Each element in statistical matrix z2
The number of times of appearance, and the result of statistics is stored in matrix w2, using w2 as the second dimensional feature vector；
By that analogy, obtain successively the sampled point that is separated by between the zero crossing of two, three, four ... zero crossings
Number, can obtain w3, w4 ...
S03, set up multidimensional characteristic matrix：
According to w1 obtained above, w2 ..., by zero padding behind wherein length short characteristic vector, make their length consistent,
The vector of a multidimensional, feature of the multidimensional characteristic vectors comprising acoustical signal may finally be obtained.The dimension of characteristic vector
It is not fixed, dimension is more, matching result can be more accurate, but amount of calculation can increases, typically takes 4.
S04, set up template base：
According to the mode of step S00 to S03,4 dimensional feature vectors are solved respectively for various different acoustical signals, build
Template base.
S05, obtain matching result：
According to the mode of step S00 to S03, the multidimensional characteristic matrix for being detected acoustical signal is extracted, and by itself and template
In storehouse, the multidimensional characteristic matrix of acoustical signal asks for Euclidean distance respectively, using Euclidean distance minimum corresponding to acoustical signal as
The acoustical signal for matching.If the signal is considered as unknown signaling more than the thresholding of setting by minima.
Cite an actual example below to illustrate said method.
By taking tank, aircraft, train and the sound whistled as an example, every kind of sound collection is twice respectively as template and survey
Sample sheet, calculates the Euclidean distance between each test sample and test template eigenmatrix, experimental result such as 1 institute of table respectively
Show, it is seen that this method can be effectively differentiated to common all kinds of acoustical signals.
1 sample of table and template matching results
In sum, presently preferred embodiments of the present invention is these are only, is not intended to limit protection scope of the present invention.
All any modification, equivalent substitution and improvements within the spirit and principles in the present invention, made etc., should be included in the present invention's
Within protection domain.
Claims (3)
1. a kind of method for recognizing sound-groove based on zero passage spaced points, it is characterised in that detailed process is：
S00, collected sound signal：
Sample rate is set as n, one section of acoustical signal is sampled, sampled point number is k；
S01, determine zero crossing：
The sampled value of all zero crossings is designated as into y (1), y (2) ... y (ε), wherein ε is the sum of all zero crossings；
S02, statistics zero passage spaced points：
First, the number of statistics adjacent zero crossing y (i+1) sampled point and y (i) between, and be stored in matrix z1, wherein
I=1,2 ... ε -1；The number of times that each element occurs in statistical matrix z1, and the result of statistics is stored in matrix w1, by w1
As the first dimensional feature vector；
Secondly, statistics is separated by the number of zero crossing y (i+2) sampled points and y (i) between of a zero crossing, and is stored to
In matrix z2, wherein i=1,2 ... ε -2；The number of times that each element occurs in statistical matrix z2, and the result storage of statistics is arrived
In matrix w2, using w2 as the second dimensional feature vector；
By that analogy, obtain successively be separated by two, three ..., the number of sampled point between the zero crossing of N-1 zero crossing,
Obtain w3, w4 ..., wN；
S03, set up multidimensional characteristic matrix：
By w1, w2 ..., the short characteristic vector subsequent Zero of length in wN, make their length consistent, obtain the characteristic vector of N-dimensional；
S04, set up template base：
According to the mode of step S00 to S03, its corresponding N-dimensional characteristic vector is obtained respectively for various different acoustical signals,
Build template base；
S05, obtain matching result：
According to the mode of step S00 to S03, the N-dimensional characteristic vector for being detected acoustical signal is extracted, and by itself and sound in template base
The characteristic vector of message number is matched, and realizes the identification of vocal print.
2. method for recognizing sound-groove according to claim 1 based on zero passage spaced points, it is characterised in that the N=4.
3. method for recognizing sound-groove according to claim 1 based on zero passage spaced points, it is characterised in that described to be matched
Process is：The N-dimensional characteristic vector of detected acoustical signal is asked for into Euclidean distance respectively with characteristic vector in template base, if minimum
Euclidean distance less than setting threshold value, using the acoustical signal corresponding to minimum euclidean distance as the acoustical signal for matching,
Otherwise, the signal being detected is considered as into unknown signaling.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510586504.0A CN105304087B (en) | 2015-09-15 | 2015-09-15 | Voiceprint recognition method based on zero-crossing separating points |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510586504.0A CN105304087B (en) | 2015-09-15 | 2015-09-15 | Voiceprint recognition method based on zero-crossing separating points |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105304087A CN105304087A (en) | 2016-02-03 |
| CN105304087B true CN105304087B (en) | 2017-03-22 |
Family
ID=55201260
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510586504.0A Active CN105304087B (en) | 2015-09-15 | 2015-09-15 | Voiceprint recognition method based on zero-crossing separating points |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105304087B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107256414A (en) * | 2017-06-27 | 2017-10-17 | 哈尔滨工业大学 | Polarimetric SAR Image convolutional neural networks sorting technique based on spatial feature collection of illustrative plates |
| CN111583938B (en) * | 2020-05-19 | 2023-02-03 | 威盛电子股份有限公司 | Electronic device and voice recognition method |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101226743A (en) * | 2007-12-05 | 2008-07-23 | 浙江大学 | Method for recognizing speaker based on conversion of neutral and affection sound-groove model |
| CN102509547A (en) * | 2011-12-29 | 2012-06-20 | 辽宁工业大学 | Method and system for voiceprint recognition based on vector quantization based |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6978238B2 (en) * | 1999-07-12 | 2005-12-20 | Charles Schwab & Co., Inc. | Method and system for identifying a user by voice |
-
2015
- 2015-09-15 CN CN201510586504.0A patent/CN105304087B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101226743A (en) * | 2007-12-05 | 2008-07-23 | 浙江大学 | Method for recognizing speaker based on conversion of neutral and affection sound-groove model |
| CN102509547A (en) * | 2011-12-29 | 2012-06-20 | 辽宁工业大学 | Method and system for voiceprint recognition based on vector quantization based |
Non-Patent Citations (1)
| Title |
|---|
| 基于过零间隔点的声纹识别技术;周家术等;《计算机工程》;20061130;第32卷(第22期);200-202 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105304087A (en) | 2016-02-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106297776B (en) | A kind of voice keyword retrieval method based on audio template | |
| US6424946B1 (en) | Methods and apparatus for unknown speaker labeling using concurrent speech recognition, segmentation, classification and clustering | |
| CN102799605B (en) | A kind of advertisement detecting method and system | |
| WO2020143263A1 (en) | Speaker identification method based on speech sample feature space trajectory | |
| CN105976809A (en) | Voice-and-facial-expression-based identification method and system for dual-modal emotion fusion | |
| Zhang et al. | A threshold-based hmm-dtw approach for continuous sign language recognition | |
| CN111462758A (en) | Method, device and equipment for intelligent conference role classification and storage medium | |
| WO2023088448A1 (en) | Speech processing method and device, and storage medium | |
| CN106250925B (en) | A kind of zero Sample video classification method based on improved canonical correlation analysis | |
| CN103531207A (en) | Voice sensibility identifying method of fused long-span sensibility history | |
| CN101447188A (en) | Digital voice print identification system and validation and identification method | |
| CN110647656B (en) | Audio retrieval method utilizing transform domain sparsification and compression dimension reduction | |
| CN105304087B (en) | Voiceprint recognition method based on zero-crossing separating points | |
| CN107564543A (en) | A kind of Speech Feature Extraction of high touch discrimination | |
| CN102073631A (en) | Video news unit dividing method by using association rule technology | |
| CN117727307B (en) | Bird voice intelligent recognition method based on feature fusion | |
| CN114038479B (en) | Method, device and storage medium for identifying and classifying bird song corresponding to low signal-to-noise ratio | |
| Somervuo | Time–frequency warping of spectrograms applied to bird sound analyses | |
| CN109344233B (en) | Chinese name recognition method | |
| Wang et al. | Automatic audio segmentation using the generalized likelihood ratio | |
| Liang et al. | Audio content classification method research based on two-step strategy | |
| CN110738987B (en) | Keyword retrieval method based on unified representation | |
| Yu | Research on music emotion classification based on CNN-LSTM network | |
| Yue et al. | Multidimensional zero-crossing interval points: a low sampling rate acoustic fingerprint recognition method | |
| Bougamouza et al. | Using Mel Frequency Cepstral Coefficient method for online Arabic characters handwriting recognition |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |