CN102750950A - Chinese emotion speech extracting and modeling method combining glottal excitation and sound track modulation information - Google Patents
Chinese emotion speech extracting and modeling method combining glottal excitation and sound track modulation information Download PDFInfo
- Publication number
- CN102750950A CN102750950A CN2012102112835A CN201210211283A CN102750950A CN 102750950 A CN102750950 A CN 102750950A CN 2012102112835 A CN2012102112835 A CN 2012102112835A CN 201210211283 A CN201210211283 A CN 201210211283A CN 102750950 A CN102750950 A CN 102750950A
- Authority
- CN
- China
- Prior art keywords
- glottal
- frequency
- speech
- emotion
- electric
- 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.)
- Granted
Links
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention relates to a Chinese emotion speech extracting and modeling method combining glottal excitation and sound track modulation information. The Chinese glottal excitation information extracting method comprises the following steps: establishing an electric glottal graph emotion speech database standard, collecting electric glottal graph emotion speech data, and subjectively evaluating the electric glottal graph emotion speech data, namely evaluating the electric glottal graph emotion speech data by ten evaluators. The Chinese emotion speech extracting and modeling method combining glottal excitation and sound track modulation information comprises the steps: firstly extracting the glottal excitation information according to an electric glottal graph; then extracting the sound track modulation information in the speech signal according to the relation of the electric glottal graph and the speech signals; and finally selecting proper emotion characteristics for every emotion by utilizing the Fisher ratios of all the optional characteristics, and training a one-to-many support vector machine. The Chinese emotion speech extracting and modeling method provided by the invention can extract the glottal excitation information through the signals of the electric glottal graph, and separate the glottal excitation information and the sound track modulation information, thereby more accurately reflecting the emotion information in voice.
Description
The present invention is that patent combines the Chinese speech sensibility extraction of glottal excitation and sound channel modulation intelligence and dividing an application of modeling method (application number 201110301307.1).
(1) technical field:
The present invention relates to a kind of Chinese speech sensibility of glottal excitation and sound channel modulation intelligence that combines and extract and modeling method, belong to areas of information technology.
(2) background technology:
The information such as affective state that include speak content, speaker ' s identity and speaker in the voice signal, speech emotional identification are the affective states that from voice signal, extracts the speaker.Voice-based excitation modulation pattern, voice signal is made up of glottal excitation and sound channel modulation two parts.Glottal excitation has determined the phonetic-rhythm variation, and identification has vital role to speech emotional.For example, when the speaker was excited, voice signal fundamental frequency average can raise, and the fundamental frequency variation range can increase; When the speaker represented query, the sentence fundamental curve part that raises up can increase.Glottal excitation has also comprised part speaker ' s identity information, and the fundamental frequency distribution range of different speakers under tranquility is different, but this difference is often fallen because of the change conductively-closed of affective state.In addition, transfer in the language having, Chinese for example, the variation of tone is also relevant with semanteme.The content of speaking that sound channel has been modulated major decision, every kind of resonance peak that vowel is corresponding different reflects different vocal tract shape information.Different speakers' vocal tract shape (sound channel length, sound channel sectional area) is distinguishing, and this difference is reflected on the effect of sound channel modulation equally.In addition, when speaker's emotional state changed, even saying same content, vocal tract shape also can change to some extent, and this variation can cause the frequency domain distribution of speech energy to change.It is thus clear that the relation between speak content, speaker ' s identity and speaker's in the voice the affective state is very complicated.
Before the present invention, the researcher mainly adopts various signal processing methods from voice signal, to extract glottal excitation and sound channel modulation intelligence, because the restriction of algorithm itself can't be accomplished separating fully of glottal excitation and sound channel modulation intelligence.
(3) summary of the invention:
The object of the invention just is to solve the defective in the above-mentioned background technology; Directly extract speaker's glottal excitation through electric glottogram signal; Utilize the convolution relation of glottal excitation and sound channel modulation to extract the sound channel modulation intelligence in the voice again; Accomplish the separation fully of the two, carry out Chinese speech sensibility identification.
A kind of Chinese speech sensibility of glottal excitation and sound channel modulation intelligence that combines of the present invention extracts and modeling method:
Wherein
The method for distilling of Chinese glottal excitation information, its step is following:
Step 1: formulate electric glottogram emotional speech normalizing database;
Step 1.1: select 20 pronunciation personnel, 20 to 30 one full year of life of age, occupation is the student, comprising 10 women, and 10 male sex;
Step 1.2: select 20 standard Chinese statements that do not have obvious emotion tendency as the recording script, average every script comprises 6.95 Chinese characters, minimum 3 words (making it snappy), maximum 12 words (I have seen your name on paper);
Step 1.3: information such as filename include file numbering, emotion classification, recording people sex, obtain manner, language classification, recording people numbering, speech sentences numbering, recording statement number of times sign;
Step 2: collect electric glottogram emotional speech data;
Step 2.1: recording location is the laboratory of a spaciousness, during recording door and window is closed, and recording time is chosen in evening; Sound pick-up outfit adopts DELL OPTIPLEX 330 computers, TIGEX-EGG3 electricity glottogram appearance, BE-8800 electret capacitor microphone; That recording software adopts is GoldWave; Record the two-channel data, L channel is the speech data that microphone is gathered, and R channel is the glottal excitation data that electric glottogram appearance is gathered; Have only operator and speaker on the scene during recording, keep indoor peace and quiet;
Step 2.2: environmental parameters such as the temperature in record speaker's personal information and recording place, humidity; And please the speaker pronounce with the descant amount; Adjustment sound card gain and microphone and speaker's mouth distance guarantee that the top can not appear cutting in speech waveform and electric glottogram waveform;
Step 2.3: please the speaker read aloud all recording scripts, repeat 3 times with the emotion of appointment;
Step 2.4: please the speaker carry out the personal evaluation, as meet my expression wish, then the result saved as sampling rate 11025Hz, the wav file of 16Bit two-channel pcm encoder form, otherwise repeating step 2.3 to recorded speech;
Step 2.5: adopt GoldWave software that the voice document that step 2.4 obtains is split, obtain 60 electric glottogram emotional speech data;
Step 2.6: please the speaker with sad, indignation, surprised, frightened, happy, detest and tranquil emotion repeating step 2.3 to step 2.5, obtain 420 electric glottogram emotional speech data;
Step 2.7: change speaker's repeating step 2.3 to step 2.6, obtain 20 speakers, comprise 10 male 10 woman, totally 8400 electric glottogram emotional speech data;
Step 3: design is based on the network emotion point evaluating system of .NET platform; Invite that evaluation and test person evaluates and tests the electric glottogram emotional speech of playing according to random sequence more than 10; According to evaluation result, concentrate selection 4 cover voice sub collection from whole voice, the voice that wherein every cover voice sub is concentrated are had at least the people of p% correctly to discern (p=70; 80,90 and 100);
In conjunction with the Chinese emotion modeling method of glottal excitation and sound channel modulation intelligence, its step is following:
Step 1: combine electric glottogram and voice signal to carry out voiced segments, voiceless sound section and quiet section division;
Step 1.1: electric glottogram signal is carried out bandpass filtering, then electric glottogram in the electric glottogram emotional speech database and voice signal are carried out the processing of branch frame, 256 sampled points of frame length, frame moves 128 sampled points;
Step 1.2: the logarithm energy E of extracting all electroacoustic door diagram data frames respectively
EggLogarithm energy E with the voice signal data frame
Speech, according to E
EggStatistics setting threshold th0, according to E
SpeechStatistics set th
1And th
2
Step 1.3: if certain Frame E
EggGreater than th
0And E
SpeechGreater than th
2, then this Frame is a voiced sound, if certain Frame is not voiced sound and E
SpeechGreater than th
1, then this Frame is a voiceless sound, the remainder data frame is quiet;
Step 2: the glottal excitation information of extracting voiced sound;
Step 2.1: calculate the frequency histogram that each voiced segments continues duration; And to taking the logarithm after the horizontal ordinate normalization of this frequency histogram; Continuing duration with voiced segments normalization logarithm is independent variable; Normalization logarithm frequency to continue duration is a dependent variable, carries out simple linear regression analysis, obtains the double-log distribution parameter that voiced segments continues the duration frequency;
Step 2.2: calculate the fundamental frequency of each unvoiced frame, calculate the double-log distribution parameter of the lasting duration frequency of fundamental curve ascent stage and descending branch according to the method for step 2.1 respectively;
Step 3: the sound channel modulation intelligence that extracts voiced sound and voiceless sound;
Step 3.1: electric glottogram and voice signal to unvoiced frame carry out FFT; Electric glottogram and voice signal to being transformed into frequency domain carry out energy normalized; Utilize frequency domain phase division; Frequency domain normalization voice signal and frequency domain normalization electricity glottogram signal are divided by; Calculate essential characteristic such as gained result's 10 rank MEL frequency cepstral coefficients, preceding 3 resonance peaks, 5 MEL sub belt energies, frequency spectrum barycenter, cutoff frequency and to first order difference and the second order difference of time, and calculate the parameters of deriving such as maximal value, minimum value, average, standard deviation, kurtosis, the degree of bias of all essential characteristics and difference characteristic;
Step 3.2: calculate essential characteristic such as 10 rank MEL frequency cepstral coefficients of voiceless sound, preceding 3 resonance peaks, 5 MEL sub belt energies, frequency spectrum barycenter, cutoff frequency and to first order difference and the second order difference of time, and calculate the parameters of deriving such as maximal value, minimum value, average, standard deviation, kurtosis, the degree of bias of all essential characteristics and difference characteristic;
Step 4: glottal excitation and sound channel modulation signature parameter according to step 1 to step 3 obtains are set up six kinds of emotion SVMs one-to-many disaggregated models;
Step 4.1: be the basis with electric glottogram emotional speech database, 6 Fisher ratios of every kind of characteristic that calculation procedure 1 to step 3 is extracted, the corresponding a kind of emotion of each Fisher ratio;
Step 4.2: the sequencing selection of the Fisher ratio that obtains according to step 4.1 is applicable to the characteristic parameter of every kind of supporting vector machine model, is every kind of emotion training one-to-many supporting vector machine model.
Good effect of the present invention and advantage are:
1. at first set up electric glottogram emotional speech database, comprised 20 speakers (10 male 10 woman), totally 8400 electric glottogram emotional speech data, and carried out subjective evaluation and test for the emotion statement in the database.
2. respectively from electric glottogram and voice signal, thoroughly separated glottal excitation and sound channel modulation intelligence, improved the robustness of speech emotional identification speak content and speaker ' s identity information.
(4) accompanying drawing subordinate list explanation:
Fig. 1 combines the extraction and the recognition methods process flow diagram of the Chinese speech sensibility point of glottal excitation harmony key system;
The oscillogram of Fig. 2 electricity glottogram and speech data;
The selected synoptic diagram of the quiet division threshold value of Fig. 3 voiceless sound voiced sound;
Fig. 4 voiced segments continues the double-log distribution parameter of duration frequency;
The double-log distribution parameter of the lasting duration frequency of Fig. 5 fundamental curve ascent stage and descending branch.
(5) embodiment:
Below in conjunction with accompanying drawing, technical scheme of the present invention is further elaborated.
As shown in Figure 1, be extraction and the recognition methods process flow diagram that combines the Chinese speech sensibility point of glottal excitation harmony key system, mainly be divided into two parts: the method for distilling of Chinese speech sensibility point and the recognition methods of Chinese speech sensibility point.
One, the method for distilling of Chinese language door excitation information, this method step is following:
Each step in the entire making process of sound bank all should be deferred to specific standard, comprises speaker standard, recording script design specifications, recording standard, audio file Naming conventions and experimental record standard etc.The making standard and the implication that are specifically related to are as shown in table 1.
Table 1
The speaker standard: select 20 pronunciation personnel, 20 to 30 one full year of life of age, occupation is the student, comprising 10 women, and 10 male sex;
1), selected statement must not comprise emotion tendency in a certain respect experiment is mainly considered following 3 aspects with the selection of recording script:; 2), must have higher emotion degree of freedom, can apply various emotions to same statement and analyze comparison; 3), be colloquial declarative sentence;
According to mentioned above principle, we have selected 20 short sentences as shown in table 2 as the recording script.
Table 2
Express speech emotional for the ease of the recording people, we have designed situational contexts and have assisted the recording people to bring out emotion.The situational contexts of recording script is directly connected to recording person's emotional expression, and its design standards is following:
1), situational contexts must meet recording person's self psychological environment or emotional experience;
2), situational contexts must have distinct emotional semantic, and be the colloquial style language of using always;
3), situational contexts is unsuitable long;
For example we have selected certain text sentence and relevant situational contexts, and are as shown in table 3 below.
Table 3
The tissue of emotional speech data and retrieval for ease has intuitively the key character of single emotional speech and to describe, and we have designed the naming rule of emotional speech file.Emotional speech file and voice primitive character file all must be named according to naming rule, and the file designation rule is as shown in table 4 below. and filename is made up of 9 characters, is divided into 7 parts.
Table 4
1), emotion classification: the emotion classification logotype occupies 1 character, forms b by first lowercase of affective style phonetic: sadness; F: indignation; J: surprised; K: fear; X: happiness; Y: detest; P: calmness.
2), recording people sex: voice signal is different with speaker's sex, and acoustic feature has evident difference.The sex sign occupies 1 character, F: women; M: the male sex.
3), obtain manner: the difference of emotional speech obtain manner makes that their effects under study for action are different, and obtain manner is divided into 2 kinds, and 0 is the recording method, and 1 is the montage method.
4), language classification: C be Chinese, E be English, for having done preparation on the later emotional speech data.
5), recording people numbering: we write down and number recording people information, and according to existing recording people scale, the recording people indicates and occupies 2 numerals, and scope is from 01 ~ 99.
6), recording text numbering: we get off the emotional speech text entry, and number, and according to the scale of existing voice text, the recording Text Flag occupies 2 numerals, and scope is from 01 ~ 99.
7), recording number of times: identified the number of times of one type of emotional expression of a recording text, occupied 1 numeral, scope is 1 ~ 9.
At first, carrying out speech emotional induces.The validity of speech emotional can be divided into nature, semi-natural and imitate 3 grades.We have adopted the following method of bringing out:
Direct given recording script and emotion classification, recording person arbitrarily associates the back recording; The director describes real script scene for recording person, then recording; The director provides corresponding recording script example, imitates the pronunciation recording after recording person listens.This several method carries out successively, and the validity of emotion can be relatively worse and worse.Therefore we begin to bring out from top first method in recording, in case can record satisfactory emotional speech, promptly finish this person's recording.If the failure would launch second method, by that analogy, be recorded to the high voice of emotion validity as far as possible.
Secondly, record; Specifically comprise the steps:
Step 2.1: recording location is the laboratory of a spaciousness, during recording door and window is closed, and recording time is chosen in evening; Sound pick-up outfit adopts DELL OPTIPLEX 330 computers, TIGEX-EGG3 electricity glottogram appearance, BE-8800 electret capacitor microphone; That recording software adopts is GoldWave; Record the two-channel data, L channel is the speech data that microphone is gathered, and R channel is the glottal excitation data that electric glottogram appearance is gathered; Have only operator and speaker on the scene during recording, keep indoor peace and quiet;
Step 2.2: environmental parameters such as the temperature in record speaker's personal information and recording place, humidity; And please the speaker pronounce with the descant amount; Adjustment sound card gain and microphone and speaker's mouth distance guarantee that the top can not appear cutting in speech waveform and electric glottogram waveform;
Step 2.3: please the speaker read aloud all recording scripts, repeat 3 times with the emotion of appointment;
Step 2.4: please the speaker carry out the personal evaluation, as meet my expression wish, then the result saved as sampling rate 11025Hz, the wav file of 16Bit two-channel pcm encoder form, otherwise repeating step 2.3 to recorded speech;
Step 2.5: adopt GoldWave software that the voice document that step 2.4 obtains is split, obtain 60 electric glottogram emotional speech data;
Step 2.6: please the speaker with sad, indignation, surprised, frightened, happy, detest and tranquil emotion repeating step 2.3 to step 2.5, obtain 420 electric glottogram emotional speech data;
Step 2.7: change speaker's repeating step 2.3 to step 2.6, obtain 20 speakers, comprise 10 male 10 woman, totally 8400 electric glottogram emotional speech data;
The subjective evaluation and test of the electric glottogram emotional speech of step 3.
Design is based on the network emotion point evaluating system of .NET platform; Invite that evaluation and test person evaluates and tests the electric glottogram emotional speech of playing according to random sequence more than 10; According to evaluation result, concentrate selection 4 cover voice sub collection from whole voice, the voice that wherein every cover voice sub is concentrated are had at least the people of p% correctly to discern (p=70; 80,90 and 100);
Two, combine the Chinese emotion modeling method of glottal excitation and sound channel modulation intelligence, this method step is following:
Step 1: combine electric glottogram and voice signal to carry out voiced segments, voiceless sound section and quiet section division;
The electricity glottogram has reflected the vibration information of vocal cords; But since speaker's swallowing act or when speaking throat's muscle movement can cause interelectrode resistance to change; This variation is also noted by electric glottogram appearance; Fig. 2 has shown the electric glottogram of male voice ", you can very great " and the oscillogram of speech data.
Step 1.1: electric glottogram signal is carried out bandpass filtering; Weaken the interference of muscular movement to electric glottogram; Then electric glottogram in the electric glottogram emotional speech database and voice signal are carried out the processing of branch frame, 256 sampled points of frame length, frame moves 128 sampled points;
Step 1.2: the logarithm energy E of extracting all electroacoustic door diagram data frames respectively
EggLogarithm energy E with the voice signal data frame
SpeechBased on the law of large numbers in the probability theory, the equal Gaussian distributed of energy of enough a large amount of unvoiced segments and sound section, as shown in Figure 3.According to E
EggStatistics setting threshold th
0, according to E
SpeechStatistics set th
1And th
2
Step 1.3: if certain Frame E
EggGreater than th
0And E
SpeechGreater than th
2, then this Frame is a voiced sound, if certain Frame is not voiced sound and E
SpeechGreater than th
1, then this Frame is a voiceless sound, the remainder data frame is quiet;
Step 2: the glottal excitation information of extracting voiced sound;
Step 2.1: calculate the frequency histogram that each voiced segments continues duration; And to taking the logarithm after the horizontal ordinate normalization of this frequency histogram; Continuing duration with voiced segments normalization logarithm is independent variable, is dependent variable with the normalization logarithm frequency that continues duration, carries out simple linear regression analysis; Obtain the double-log distribution parameter that voiced segments continues the duration frequency, as shown in Figure 4;
Step 2.2: calculate the fundamental frequency of each unvoiced frame, the method that accordings to step 2.1 is calculated the double-log distribution parameter of the lasting duration frequency of fundamental curve ascent stage and descending branch respectively, and is as shown in Figure 5;
Step 3: the sound channel modulation intelligence that extracts voiced sound and voiceless sound;
Step 3.1: electric glottogram and voice signal to unvoiced frame carry out FFT; Electric glottogram and voice signal to being transformed into frequency domain carry out energy normalized; Utilize frequency domain phase division; Frequency domain normalization voice signal and frequency domain normalization electricity glottogram signal are divided by; Calculate essential characteristic such as gained result's 10 rank MEL frequency cepstral coefficients, preceding 3 resonance peaks, 5 MEL sub belt energies, frequency spectrum barycenter, cutoff frequency and to first order difference and the second order difference of time, and calculate the parameters of deriving such as maximal value, minimum value, average, standard deviation, kurtosis, the degree of bias of all essential characteristics and difference characteristic;
Fundamental frequency, energy, zero-crossing rate, resonance peak all belong to traditional phonetic feature, wherein the low frequency rhythm characteristic of energy and fundamental frequency reaction voice, zero-crossing rate and resonance peak reaction high-frequency information.
Frequency spectrum barycenter and frequency spectrum cutoff frequency have reflected the frequency distribution characteristic of voice signal, all belong to the tonequality characteristic.At first every frame data are carried out quick Fourier transformation:
Wherein F (k) representes the amplitude of k Frequency point, and x (n) representes input speech signal, and N is voice signal frame length (1024).Frequency spectrum barycenter F
CentBe defined as:
Frequency spectrum cutoff frequency F
CutBe meant a certain Frequency point, satisfy:
Step 3.2: calculate essential characteristic such as 10 rank MEL frequency cepstral coefficients of voiceless sound, preceding 3 resonance peaks, 5 MEL sub belt energies, frequency spectrum barycenter, cutoff frequency and to first order difference and the second order difference of time, and calculate the parameters of deriving such as maximal value, minimum value, average, standard deviation, kurtosis, the degree of bias of all essential characteristics and difference characteristic;
Step 4: glottal excitation and sound channel modulation signature parameter according to step 1 to step 3 obtains are set up six kinds of emotion SVMs one-to-many disaggregated models;
The Fisher criterion is a kind of traditional linear discriminant method, and its ultimate principle is to seek certain projection subspace of feature space, makes all unique points obtain best classification in this subspace.If sample set X={x
1, x
2..., x
nIn total n sample, branch C class, each sample x
iBe the L dimensional vector, dispersion S between the definition class
bWith dispersion S in the class
wBe respectively:
X wherein
j, n
jAnd m
jBe respectively the subclass of j class sample, total number and average point, m is the average point of all samples.
The Fisher ratio is defined as:
F
r=diag(S
b./S
w) (6)
Wherein the corresponding element of symbol " ./" representing matrix is divided by respectively, and function d iag gets the element on the diagonal of a matrix.F
rBe the L dimensional vector, wherein each element has reflected two classification capacities of character pair.
Step 4.1: be the basis with electric glottogram emotional speech database, 6 Fisher ratios of every kind of characteristic that calculation procedure 1 to step 3 is extracted, the corresponding a kind of emotion of each Fisher ratio;
Step 4.2: the sequencing selection of the Fisher ratio that obtains according to step 4.1 is applicable to the characteristic parameter of every kind of supporting vector machine model, is every kind of emotion training one-to-many supporting vector machine model.
Claims (1)
1. combine the Chinese emotion modeling method of glottal excitation and sound channel modulation intelligence, its step is following:
Step 1: combine electric glottogram and voice signal to carry out voiced segments, voiceless sound section and quiet section division;
Step 1.1: electric glottogram signal is carried out bandpass filtering, then electric glottogram in the electric glottogram emotional speech database and voice signal are carried out the processing of branch frame, 256 sampled points of frame length, frame moves 128 sampled points;
Step 1.2: the logarithm energy E of extracting all electroacoustic door diagram data frames respectively
EggLogarithm energy E with the voice signal data frame
Speech, according to E
EggStatistics setting threshold th
0, according to E
SpeechStatistics set th
1And th
2
Step 1.3: if certain Frame E
EggGreater than th
0And E
SpeechGreater than th
2, then this Frame is a voiced sound, if certain Frame is not voiced sound and E
SpeechGreater than th
1, then this Frame is a voiceless sound, the remainder data frame is quiet;
Step 2: the glottal excitation information of extracting voiced sound;
Step 2.1: calculate the frequency histogram that each voiced segments continues duration; And to taking the logarithm after the horizontal ordinate normalization of this frequency histogram; Continuing duration with voiced segments normalization logarithm is independent variable; Normalization logarithm frequency to continue duration is a dependent variable, carries out simple linear regression analysis, obtains the double-log distribution parameter that voiced segments continues the duration frequency;
Step 2.2: calculate the fundamental frequency of each unvoiced frame, calculate the double-log distribution parameter of the lasting duration frequency of fundamental curve ascent stage and descending branch according to the method for step 2.1 respectively;
Step 3: the sound channel modulation intelligence that extracts voiced sound and voiceless sound;
Step 3.1: electric glottogram and voice signal to unvoiced frame carry out FFT; Electric glottogram and voice signal to being transformed into frequency domain carry out energy normalized; Utilize frequency domain phase division; Frequency domain normalization voice signal and frequency domain normalization electricity glottogram signal are divided by; Calculate essential characteristic such as gained result's 10 rank MEL frequency cepstral coefficients, preceding 3 resonance peaks, 5 MEL sub belt energies, frequency spectrum barycenter, cutoff frequency and to first order difference and the second order difference of time, and calculate the parameters of deriving such as maximal value, minimum value, average, standard deviation, kurtosis, the degree of bias of all essential characteristics and difference characteristic;
Step 3.2: calculate essential characteristic such as 10 rank MEL frequency cepstral coefficients of voiceless sound, preceding 3 resonance peaks, 5 MEL sub belt energies, frequency spectrum barycenter, cutoff frequency and to first order difference and the second order difference of time, and calculate the parameters of deriving such as maximal value, minimum value, average, standard deviation, kurtosis, the degree of bias of all essential characteristics and difference characteristic;
Step 4: glottal excitation and sound channel modulation signature parameter according to step 1 to step 3 obtains are set up six kinds of emotion SVMs one-to-many disaggregated models;
Step 4.1: be the basis with electric glottogram emotional speech database, 6 Fisher ratios of every kind of characteristic that calculation procedure 1 to step 3 is extracted, the corresponding a kind of emotion of each Fisher ratio;
Step 4.2: the sequencing selection of the Fisher ratio that obtains according to step 4.1 is applicable to the characteristic parameter of every kind of supporting vector machine model, is every kind of emotion training one-to-many supporting vector machine model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201210211283.5A CN102750950B (en) | 2011-09-30 | 2011-09-30 | Chinese emotion speech extracting and modeling method combining glottal excitation and sound track modulation information |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201210211283.5A CN102750950B (en) | 2011-09-30 | 2011-09-30 | Chinese emotion speech extracting and modeling method combining glottal excitation and sound track modulation information |
Related Parent Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN 201110301307 Division CN102411932B (en) | 2011-09-30 | 2011-09-30 | Methods for extracting and modeling Chinese speech emotion in combination with glottis excitation and sound channel modulation information |
Publications (2)
Publication Number | Publication Date |
---|---|
CN102750950A true CN102750950A (en) | 2012-10-24 |
CN102750950B CN102750950B (en) | 2014-04-16 |
Family
ID=47031072
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201210211283.5A Expired - Fee Related CN102750950B (en) | 2011-09-30 | 2011-09-30 | Chinese emotion speech extracting and modeling method combining glottal excitation and sound track modulation information |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN102750950B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104409081A (en) * | 2014-11-25 | 2015-03-11 | 广州酷狗计算机科技有限公司 | Speech signal processing method and device |
CN104732981A (en) * | 2015-03-17 | 2015-06-24 | 北京航空航天大学 | Voice annotation method for Chinese speech emotion database combined with electroglottography |
CN104851421A (en) * | 2015-04-10 | 2015-08-19 | 北京航空航天大学 | Voice processing method and device |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1391876A1 (en) * | 2002-08-14 | 2004-02-25 | Sony International (Europe) GmbH | Method of determining phonemes in spoken utterances suitable for recognizing emotions using voice quality features |
CN101261832A (en) * | 2008-04-21 | 2008-09-10 | 北京航空航天大学 | Extraction and modeling method for Chinese speech sensibility information |
CN101620852A (en) * | 2008-07-01 | 2010-01-06 | 邹采荣 | Speech-emotion recognition method based on improved quadratic discriminant |
US20100204990A1 (en) * | 2008-09-26 | 2010-08-12 | Yoshifumi Hirose | Speech analyzer and speech analysys method |
CN101930735A (en) * | 2009-06-23 | 2010-12-29 | 富士通株式会社 | Speech emotion recognition equipment and speech emotion recognition method |
-
2011
- 2011-09-30 CN CN201210211283.5A patent/CN102750950B/en not_active Expired - Fee Related
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1391876A1 (en) * | 2002-08-14 | 2004-02-25 | Sony International (Europe) GmbH | Method of determining phonemes in spoken utterances suitable for recognizing emotions using voice quality features |
CN101261832A (en) * | 2008-04-21 | 2008-09-10 | 北京航空航天大学 | Extraction and modeling method for Chinese speech sensibility information |
CN101620852A (en) * | 2008-07-01 | 2010-01-06 | 邹采荣 | Speech-emotion recognition method based on improved quadratic discriminant |
US20100204990A1 (en) * | 2008-09-26 | 2010-08-12 | Yoshifumi Hirose | Speech analyzer and speech analysys method |
CN101930735A (en) * | 2009-06-23 | 2010-12-29 | 富士通株式会社 | Speech emotion recognition equipment and speech emotion recognition method |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104409081A (en) * | 2014-11-25 | 2015-03-11 | 广州酷狗计算机科技有限公司 | Speech signal processing method and device |
CN104732981A (en) * | 2015-03-17 | 2015-06-24 | 北京航空航天大学 | Voice annotation method for Chinese speech emotion database combined with electroglottography |
CN104732981B (en) * | 2015-03-17 | 2018-01-12 | 北京航空航天大学 | A kind of voice annotation method of the Chinese speech sensibility database of combination ElectroglottographicWaveform |
CN104851421A (en) * | 2015-04-10 | 2015-08-19 | 北京航空航天大学 | Voice processing method and device |
CN104851421B (en) * | 2015-04-10 | 2018-08-17 | 北京航空航天大学 | Method of speech processing and device |
Also Published As
Publication number | Publication date |
---|---|
CN102750950B (en) | 2014-04-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN102411932B (en) | Methods for extracting and modeling Chinese speech emotion in combination with glottis excitation and sound channel modulation information | |
CN101261832B (en) | Extraction and modeling method for Chinese speech sensibility information | |
CN102893326B (en) | Chinese voice emotion extraction and modeling method combining emotion points | |
Kandali et al. | Emotion recognition from Assamese speeches using MFCC features and GMM classifier | |
CN102655003B (en) | Method for recognizing emotion points of Chinese pronunciation based on sound-track modulating signals MFCC (Mel Frequency Cepstrum Coefficient) | |
CN102332263B (en) | Close neighbor principle based speaker recognition method for synthesizing emotional model | |
Pao et al. | Mandarin emotional speech recognition based on SVM and NN | |
Shaw et al. | Emotion recognition and classification in speech using artificial neural networks | |
Selvaraj et al. | Human speech emotion recognition | |
Matin et al. | A speech emotion recognition solution-based on support vector machine for children with autism spectrum disorder to help identify human emotions | |
Chauhan et al. | Speech to text converter using Gaussian Mixture Model (GMM) | |
Verma et al. | Age driven automatic speech emotion recognition system | |
CN114550706B (en) | Intelligent campus voice recognition method based on deep learning | |
Farooq et al. | Mispronunciation detection in articulation points of Arabic letters using machine learning | |
Qadri et al. | A critical insight into multi-languages speech emotion databases | |
Mary et al. | Analysis and detection of mimicked speech based on prosodic features | |
CN102750950B (en) | Chinese emotion speech extracting and modeling method combining glottal excitation and sound track modulation information | |
Haque et al. | Modification of energy spectra, epoch parameters and prosody for emotion conversion in speech | |
Chaudhari et al. | Methodology for gender identification, classification and recognition of human age | |
Bansal et al. | Emotional Hindi speech: Feature extraction and classification | |
Gautam et al. | Developmental pattern analysis and age prediction by extracting speech features and applying various classification techniques | |
Rahman et al. | Development of automated Tajweed checking system for children in learning Quran | |
Lee et al. | Speech emotion recognition using spectral entropy | |
Dumpala et al. | Analysis of the Effect of Speech-Laugh on Speaker Recognition System. | |
Rupasinghe et al. | Robust Speech Analysis Framework Using CNN |
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 | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20140416 Termination date: 20140930 |
|
EXPY | Termination of patent right or utility model |