KR101671305B1 - Apparatus for extracting feature parameter of input signal and apparatus for recognizing speaker using the same - Google Patents
Apparatus for extracting feature parameter of input signal and apparatus for recognizing speaker using the same Download PDFInfo
- Publication number
- KR101671305B1 KR101671305B1 KR1020150183897A KR20150183897A KR101671305B1 KR 101671305 B1 KR101671305 B1 KR 101671305B1 KR 1020150183897 A KR1020150183897 A KR 1020150183897A KR 20150183897 A KR20150183897 A KR 20150183897A KR 101671305 B1 KR101671305 B1 KR 101671305B1
- Authority
- KR
- South Korea
- Prior art keywords
- signal
- input signal
- feature parameter
- pitch
- unit
- Prior art date
Links
- 230000000737 periodic effect Effects 0.000 claims abstract description 57
- 230000005284 excitation Effects 0.000 claims abstract description 47
- 238000001228 spectrum Methods 0.000 claims abstract description 16
- 238000000034 method Methods 0.000 claims description 32
- 238000005311 autocorrelation function Methods 0.000 claims description 24
- 238000012545 processing Methods 0.000 claims description 16
- 238000000605 extraction Methods 0.000 claims description 14
- 238000001514 detection method Methods 0.000 claims description 11
- 238000001914 filtration Methods 0.000 claims description 7
- 238000007781 pre-processing Methods 0.000 claims description 4
- 230000001131 transforming effect Effects 0.000 claims description 2
- 239000000284 extract Substances 0.000 description 7
- 238000004364 calculation method Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 230000000052 comparative effect Effects 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- 210000001260 vocal cord Anatomy 0.000 description 2
- 238000004891 communication Methods 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
-
- 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
- G10L17/00—Speaker identification or verification techniques
- G10L17/02—Preprocessing operations, e.g. segment selection; Pattern representation or modelling, e.g. based on linear discriminant analysis [LDA] or principal components; Feature selection or extraction
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- Algebra (AREA)
- Mathematical Analysis (AREA)
- Pure & Applied Mathematics (AREA)
- Computational Mathematics (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Optimization (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
Description
The present invention relates to an apparatus for extracting characteristic parameters of an input signal and an apparatus for recognizing a speaker using the apparatus.
Speech processing technology, which computer processes and understands human speech, is a promising technology that can be used in various fields. In particular, a speaker recognition technique for identifying a speaker based on the input voice may be used for identity verification in a security system or for user identification in an intelligent robot.
In general, speech recognition including speaker recognition extracts a feature vector from a speech input signal and compares it with previously stored data to recognize the information. However, the speech recognition rate of the present technology is limited to be commercialized in various fields, and among them, the recognition rate of the speaker recognition technology is not high, and continuous research and development is necessary.
An object of the present invention is to provide a feature parameter extracting apparatus capable of improving the speaker recognition rate and a speaker recognizing apparatus using the feature parameter extracting apparatus.
An apparatus for extracting feature parameters according to an exemplary embodiment of the present invention includes: a periodic signal detector for detecting a periodic interval of an input signal; An excitation signal extractor for extracting an excitation signal in a periodic interval of the input signal; And a feature parameter calculation unit for calculating a feature parameter that characterizes the input signal based on the frequency response spectrum of the excitation signal.
The periodic signal detector may detect a periodic interval based on a result of an auto-correlation function for the input signal.
The periodic signal detector may determine that the periodic signal has a period T when the input signal is x (n) and R (T) / R (0) for the input signal is greater than or equal to a predetermined threshold, R (T) =? X (n) x (n + T).
Wherein the characteristic parameter extracting unit comprises: a representative pitch detecting unit for determining a representative pitch value among pitch values of the input signal; And a proper pitch detector for determining an appropriate pitch value based on the representative pitch value among the pitch values in the periodic interval of the input signal, wherein the feature parameter extractor extracts the feature parameter Can be calculated.
The representative pitch detection unit may determine the median value as the representative pitch value by arranging the pitch values of the input signals in order of magnitude.
The appropriate pitch detector may determine a value that is closest to the representative pitch value among the pitch values selected based on the result of the autocorrelation function for the input signal of the periodic interval as a proper pitch value.
The excitation signal extraction unit may include a line emphasis unit that preprocesses the input signal of the periodic section and outputs a preprocessed signal in which high frequency components lost in the process of generating the input signal are compensated.
Wherein the excitation signal extracting unit comprises: an autocorrelation function estimator for outputting a result of an autocorrelation function for the preprocessed signal; A speculative coefficient calculator receiving an output of the autocorrelation function estimator and outputting a speculative coefficient based on a Levinson-Durbin algorithm; And an inverse filtering unit performing inverse filtering based on the preprocessing signal and the estimated coefficient to output the excitation signal.
The feature parameter calculator may include a frequency domain transformer for performing a discrete Fourier transform on the excitation signal based on the predetermined pitch value to convert the excitation signal into a discrete Fourier spectrum in the frequency domain.
The feature parameter calculator may calculate the logarithm of the magnitude of the discrete Fourier spectra with the feature parameter.
Wherein the feature parameter calculating unit calculates a feature frequency of the input signal based on a Mel-Frequency (M-F) filter that applies a Mel-Frequency filter to a frequency response spectrum of the input signal with respect to a periodic interval and an aperiodic interval of the input signal, A response obtaining unit; And a cepstral coefficient acquiring unit for performing an inverse discrete cosine transform of the mel-frequency response to obtain a cepstrum coefficient.
The feature parameter calculating unit may calculate a value obtained by multiplying the output of the feature parameter calculating unit and the output of the cepstrum coefficient acquiring unit by predetermined weights, respectively, as a feature parameter.
A speaker recognition apparatus according to an embodiment of the present invention includes: a voice collection unit for collecting a voice of a speaker; A voice processing unit for processing the collected voice and discriminating whether or not it matches the voice of a previously registered user; And a storage unit for storing information on the voice of the user.
The voice processing unit may include: a periodic signal detector for detecting a period of a signal of the input signal; An excitation signal extractor for extracting an excitation signal in a periodic interval of the input signal; And a feature parameter calculation unit for calculating a feature parameter that characterizes the input signal based on the frequency response spectrum of the excitation signal.
According to an embodiment of the present invention, a feature parameter extracting apparatus that increases the speech recognition rate in speech signal processing can be obtained, and a speaker recognition apparatus improved in speaker recognition rate can be obtained.
1 is an exemplary block diagram of a speaker recognition apparatus according to an embodiment of the present invention.
2 is an exemplary block diagram of a feature parameter extraction unit according to an embodiment of the present invention.
3 is an exemplary flowchart of a method of detecting a representative pitch in the feature parameter extracting unit of FIG.
4A and 4B are graphs for explaining a method of detecting a representative pitch in an input signal according to an embodiment of the present invention.
5 is an analysis graph of an autocorrelation function for the periodic signal detected by the periodic signal detecting unit of the characteristic parameter extracting unit of FIG.
FIG. 6 is an exemplary flowchart of a method for detecting an appropriate pitch in the feature parameter extracting unit of FIG. 2. FIG.
7 is a graph for explaining a method of detecting a proper pitch in an input signal according to an embodiment of the present invention.
8 is an exemplary flowchart of a method of extracting an excitation signal in the feature parameter calculating unit of FIG.
9 is a table showing improved speaker recognition rates according to an embodiment of the present invention.
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings attached hereto.
1 is an exemplary block diagram of a
1, the
The
The
The
According to an embodiment of the present invention, the
2 is an exemplary block diagram of a feature parameter extraction unit according to an embodiment of the present invention.
2, the feature
The feature
3 is an exemplary flowchart of a method of detecting a representative pitch in the representative pitch detecting unit 12131 included in the characteristic
Referring to FIG. 3, the representative pitch detector 12131 detects a pitch by analyzing an autocorrelation function. The autocorrelation function is a function indicating the correlation of a value taken by two arbitrary time points of an arbitrary signal. For example, it is possible to show a correlation between values taken at time t + τ delayed by a predetermined time from t-time and t-time. The pitch can be detected at a position where the autocorrelation function of the speech signal becomes the maximum. Specifically, it can be detected by the following equation.
r x (t) = Σx (n) x (n + τ), (τ = nT 0 , r x (nT 0 ) = r x
n is an integer, and the fundamental period of the periodic signal is defined as T 0 .
Referring back to FIG. 3, the representative pitch detector 12131 may sort the pitch values of the input signals in order of magnitude, and determine the median value as the representative pitch value.
4A and 4B are graphs for explaining a method of detecting a representative pitch in an input signal according to an embodiment of the present invention.
4A is a distribution of pitch values detected from an exemplary input signal. As shown in FIG. 4A, the pitch value may suddenly increase (pitch doubling) or suddenly decrease (pitch haven) may occur. If the pitch value is used to perform speech recognition, the speech recognition rate and the speaker recognition rate Can be lowered. Accordingly, in order to correct such a pitch extraction error, the representative pitch detector 12131 can sort the pitch values in order of magnitude and determine the median value as the representative pitch value as shown in FIG. 4B.
The representative pitch detection unit 12131 can determine a representative pitch of the speaker for an arbitrary speaker and correct the pitch value used in future speech recognition and speaker recognition based on the representative pitch value. That is, the pitch extraction error in the speech frame to be analyzed can be corrected using the representative pitch value. The determined representative pitch value is used to determine an appropriate pitch value in the analysis of the periodic signal. The related contents will be described in detail with reference to FIGS. 6 and 7. FIG.
Referring back to FIG. 2, the
A voice signal can be input as an input signal. The voices are divided into voiced voices that cause vibrations in the vocal cords and unvoiced voices that do not cause vibrations of the vocal cords. Periodic excitation signals can be detected in voiced intervals. The apparatus for extracting feature parameters according to an embodiment of the present invention detects feature parameters from excitation signals and uses the feature parameters as auxiliary parameters for improving the speech recognition rate. Therefore, it is necessary to divide the voiced part of the input signal, that is, the period of the periodic signal.
According to an embodiment of the present invention, the
5 is an analysis graph of an autocorrelation function for the periodic signal detected by the periodic signal detecting unit of the characteristic parameter extracting unit of FIG. The
Referring again to FIG. 2, the
Referring to FIG. 6, the optimum pitch detector 12132 may estimate an autocorrelation function for an input signal and analyze an outline of an autocorrelation function graph to determine a proper pitch value. 7 is a graph for explaining a method of detecting a proper pitch in an input signal according to an embodiment of the present invention. The appropriate pitch detection unit 12132 can select candidate pitch values based on the autocorrelation function. As shown in FIG. 2, the optimum pitch detection unit 12132 can determine a proper pitch value based on the great pitch value among the candidate pitch values. The appropriate pitch detection unit 12132 can determine a value that is closest to the representative pitch value among the candidate pitch values as an appropriate pitch value. By using a value that is closest to the representative pitch value as the optimum pitch value, the possibility of pitch doubling and pitch havening phenomenon as described above can be reduced.
8 is an exemplary flowchart of a method of extracting an excitation signal in the excitation
Referring to FIG. 8, the excitation
H (z) = 1 -? Z -1 , (0.9?? 1 )
Typically a = 0.97.
8, the excitation
The inverse filtering unit may be implemented to have a transfer function expressed by the following equation.
Here, P may be an appropriate pitch value detected by the appropriate pitch detecting unit 12132. [ If the appropriate pitch value detected by the appropriate pitch detecting unit 12132 is not an integer, the integer P value can be calculated through a process such as rounding.
According to an embodiment of the present invention, the
The feature
, Where p may be the pitch value detected in the appropriate pitch detecting section. If the pitch value detected by the appropriate pitch detector is not an integer, the rounded value may be p.
The excitation signal e (n) is subjected to discrete Fourier transform to obtain a discrete Fourier spectrum for the excitation signal.
The excitation signal e (n) can be transformed as follows.
Here, the feature parameters that characterize the speech signal can be calculated using the DFS coefficients A k and B k .
In one embodiment, the
According to one embodiment, the feature
Where K is the harmonic number of the pitch value and E k is the magnitude of the Kth DFS coefficient extracted from the excitation signal normalized by the energy of the input signal.
Referring again to FIG. 2, the feature
According to one embodiment, the
The feature
Referring again to FIG. 1, in the
According to one embodiment, the
In one embodiment, the score may be calculated as:
x t is the feature vector in the analyzed speech frame, and S i is the Gaussian mixture model parameter for the particular speaker. P represents the probability that the feature vector x t will occur in the speaker's sound.
According to an embodiment of the present invention, the score may be calculated by applying various parameters extracted by the feature
9 is a table showing improved speaker recognition rates according to an embodiment of the present invention. Referring to FIG. 9, speaker recognition rates are shown when speaker recognition is performed by calculating a score according to an embodiment of the present invention, and when speaker recognition is performed by calculating scores according to a comparative example. In the case of the embodiment, both the MFCC parameter and the DFS coefficient parameter of the excitation signal are used. In the case of the comparative example, only the MFCC parameter is used. The score weights?,?, And? Represent weights assigned to the scores calculated from the MFCC extracted from the periodic signal, the MFCC extracted from the aperiodic signal, and the DFS coefficient parameters extracted from the excitation signal.
As shown in the table of FIG. 9, according to an embodiment of the present invention, the speaker recognition rate is further improved when additional parameters extracted from the excitation signal are used.
While the present invention has been described with reference to the exemplary embodiments, it is to be understood that the invention is not limited to the disclosed exemplary embodiments. Those skilled in the art will appreciate that various modifications may be made to the embodiments described above. The scope of the present invention is defined only by the interpretation of the appended claims.
10: Speaker recognition device
110:
120:
121: Feature parameter extracting unit
130:
1211: Periodic signal detection unit
1212:
1213: Feature parameter calculating section
Claims (13)
An excitation signal extractor for extracting an excitation signal in a periodic interval of the input signal;
A feature parameter calculating unit for calculating a feature parameter that characterizes the input signal based on a frequency response spectrum of the excitation signal;
A representative pitch detector for determining a representative pitch value among pitch values of the input signal; And
And an appropriate pitch detector for determining an appropriate pitch value based on the representative pitch value among the pitch values in the periodic interval of the input signal,
Wherein the feature parameter extraction unit calculates the feature parameter based on the appropriate pitch value.
Wherein the periodic signal detection unit comprises:
And detects a periodic interval based on a result of an auto-correlation function on the input signal.
Wherein the periodic signal detection unit comprises:
(T) / R (0) for the input signal is equal to or greater than a predetermined threshold value,
Wherein R (T) = Sigmax (n) x (n + T).
The representative pitch detecting unit may include:
And arranging the pitch values of the input signals in order of magnitude to determine an intermediate value as the representative pitch value.
The optimum pitch detection unit may include:
And determines a value that is closest to the representative pitch value among the pitch values selected based on the result of the autocorrelation function for the input signal of the periodic interval as a proper pitch value.
Wherein the excitation signal extracting unit comprises:
And a line emphasis unit for preprocessing the input signal of the periodic section and outputting a preprocessed signal compensated for high frequency components lost in the process of generating the input signal.
Wherein the excitation signal extracting unit comprises:
An autocorrelation function output unit for outputting the result of the autocorrelation function for the preprocessed signal; And
A speculative coefficient calculator receiving an output of the autocorrelation function estimator and outputting a speculative coefficient based on a Levinson-Durbin algorithm; And
And an inverse filtering unit performing inverse filtering on the basis of the pre-processing signal and the estimated coefficient to output the excitation signal.
Wherein the feature parameter calculating unit comprises:
And a frequency domain transformer for transforming the excitation signal into a discrete Fourier spectrum in a frequency domain by performing a discrete Fourier transform on the basis of the appropriate pitch value.
Wherein the feature parameter calculating unit comprises:
And the logarithm of the size of the discrete Fourier spectrum is calculated as the feature parameter.
Wherein the feature parameter calculating unit comprises:
And a value obtained by adding predetermined weights to the energy of the order of the discrete Fourier spectrum and calculating a sum as a feature parameter.
Wherein the feature parameter extracting unit comprises:
A mel-frequency response acquiring unit for acquiring a mel-frequency response by applying a Mel-Frequency filter to a frequency response spectrum of the input signal with respect to a periodic interval and an aperiodic interval of the input signal; And
And a cepstrum coefficient obtaining unit for obtaining a cepstrum coefficient by inverse discrete cosine transform of the Mel-frequency response.
A voice processing unit for processing the collected voice and discriminating whether or not it matches the voice of a previously registered user; And
And a storage unit for storing information on the voice of the user,
Wherein the voice processing unit comprises:
A periodic signal detector for detecting a periodic interval of an input signal;
An excitation signal extractor for extracting an excitation signal in a periodic interval of the input signal;
A feature parameter calculating unit for calculating a feature parameter that characterizes the input signal based on a frequency response spectrum of the excitation signal;
A representative pitch detector for determining a representative pitch value among the pitch values of the input signal; And
And an appropriate pitch detector for determining an appropriate pitch value based on the representative pitch value among the pitch values in the periodic interval of the input signal,
Wherein the feature parameter extraction unit calculates the feature parameter based on the appropriate pitch value.
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
KR1020150183897A KR101671305B1 (en) | 2015-12-22 | 2015-12-22 | Apparatus for extracting feature parameter of input signal and apparatus for recognizing speaker using the same |
PCT/KR2016/014673 WO2017111386A1 (en) | 2015-12-22 | 2016-12-14 | Apparatus for extracting feature parameters of input signal, and speaker recognition apparatus using same |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
KR1020150183897A KR101671305B1 (en) | 2015-12-22 | 2015-12-22 | Apparatus for extracting feature parameter of input signal and apparatus for recognizing speaker using the same |
Publications (1)
Publication Number | Publication Date |
---|---|
KR101671305B1 true KR101671305B1 (en) | 2016-11-02 |
Family
ID=57518247
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
KR1020150183897A KR101671305B1 (en) | 2015-12-22 | 2015-12-22 | Apparatus for extracting feature parameter of input signal and apparatus for recognizing speaker using the same |
Country Status (2)
Country | Link |
---|---|
KR (1) | KR101671305B1 (en) |
WO (1) | WO2017111386A1 (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108053837A (en) * | 2017-12-28 | 2018-05-18 | 深圳市保千里电子有限公司 | A kind of method and system of turn signal voice signal identification |
KR20220065343A (en) | 2020-11-13 | 2022-05-20 | 서울시립대학교 산학협력단 | Apparatus for simultaneously performing spoofing attack detection and speaker recognition based on deep neural network and method therefor |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113012716B (en) * | 2021-02-26 | 2023-08-04 | 武汉星巡智能科技有限公司 | Infant crying type identification method, device and equipment |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20050048214A (en) * | 2003-11-19 | 2005-05-24 | 학교법인연세대학교 | Method and system for pith synchronous feature generation of speaker recognition system |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7684988B2 (en) * | 2004-10-15 | 2010-03-23 | Microsoft Corporation | Testing and tuning of automatic speech recognition systems using synthetic inputs generated from its acoustic models |
KR100933946B1 (en) * | 2007-10-29 | 2009-12-28 | 연세대학교 산학협력단 | Feature vector extraction method using adaptive selection of frame shift and speaker recognition system thereof |
KR20100036893A (en) * | 2008-09-30 | 2010-04-08 | 삼성전자주식회사 | Speaker cognition device using voice signal analysis and method thereof |
-
2015
- 2015-12-22 KR KR1020150183897A patent/KR101671305B1/en active IP Right Grant
-
2016
- 2016-12-14 WO PCT/KR2016/014673 patent/WO2017111386A1/en active Application Filing
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20050048214A (en) * | 2003-11-19 | 2005-05-24 | 학교법인연세대학교 | Method and system for pith synchronous feature generation of speaker recognition system |
Non-Patent Citations (1)
Title |
---|
강지훈, 정상배, ‘유/무성음 구분 및 이중적 특징파라미터 결합을 이용한 화자인식 성능 개선’, 한국정보통신학회논문지, Vol.18, No.6, pp.1294~1301, June 2014.* * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108053837A (en) * | 2017-12-28 | 2018-05-18 | 深圳市保千里电子有限公司 | A kind of method and system of turn signal voice signal identification |
KR20220065343A (en) | 2020-11-13 | 2022-05-20 | 서울시립대학교 산학협력단 | Apparatus for simultaneously performing spoofing attack detection and speaker recognition based on deep neural network and method therefor |
Also Published As
Publication number | Publication date |
---|---|
WO2017111386A1 (en) | 2017-06-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106935248B (en) | Voice similarity detection method and device | |
Tiwari | MFCC and its applications in speaker recognition | |
Sahidullah et al. | A comparison of features for synthetic speech detection | |
US9536547B2 (en) | Speaker change detection device and speaker change detection method | |
CN108281146B (en) | Short voice speaker identification method and device | |
Rakesh et al. | Gender Recognition using speech processing techniques in LABVIEW | |
US20130035933A1 (en) | Audio signal processing apparatus and audio signal processing method | |
CN108305639B (en) | Speech emotion recognition method, computer-readable storage medium and terminal | |
Vyas | A Gaussian mixture model based speech recognition system using Matlab | |
CN108682432B (en) | Speech emotion recognition device | |
Chaudhary et al. | Gender identification based on voice signal characteristics | |
WO2018095167A1 (en) | Voiceprint identification method and voiceprint identification system | |
KR101671305B1 (en) | Apparatus for extracting feature parameter of input signal and apparatus for recognizing speaker using the same | |
AboElenein et al. | Improved text-independent speaker identification system for real time applications | |
Srinivas et al. | Relative phase shift features for replay spoof detection system | |
EP0474496B1 (en) | Speech recognition apparatus | |
Singh et al. | Novel feature extraction algorithm using DWT and temporal statistical techniques for word dependent speaker’s recognition | |
Maazouzi et al. | MFCC and similarity measurements for speaker identification systems | |
Sadjadi et al. | Robust front-end processing for speaker identification over extremely degraded communication channels | |
Kaminski et al. | Automatic speaker recognition using a unique personal feature vector and Gaussian Mixture Models | |
Kumar et al. | Effective preprocessing of speech and acoustic features extraction for spoken language identification | |
Natarajan et al. | Segmentation of continuous Tamil speech into syllable like units | |
US20090063149A1 (en) | Speech retrieval apparatus | |
JP4537821B2 (en) | Audio signal analysis method, audio signal recognition method using the method, audio signal section detection method, apparatus, program and recording medium thereof | |
Sharma et al. | Speech recognition of Punjabi numerals using synergic HMM and DTW approach |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
GRNT | Written decision to grant | ||
FPAY | Annual fee payment |
Payment date: 20190930 Year of fee payment: 4 |