CN108735232A - A kind of personality recognition methods and device - Google Patents
A kind of personality recognition methods and device Download PDFInfo
- Publication number
- CN108735232A CN108735232A CN201710271654.1A CN201710271654A CN108735232A CN 108735232 A CN108735232 A CN 108735232A CN 201710271654 A CN201710271654 A CN 201710271654A CN 108735232 A CN108735232 A CN 108735232A
- Authority
- CN
- China
- Prior art keywords
- piece
- voice sub
- feature information
- acoustic feature
- emotion
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 41
- 230000008451 emotion Effects 0.000 claims abstract description 79
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 46
- 238000005259 measurement Methods 0.000 claims description 44
- 230000002996 emotional effect Effects 0.000 claims description 43
- 230000036651 mood Effects 0.000 claims description 42
- 230000006870 function Effects 0.000 claims description 29
- 206010022998 Irritability Diseases 0.000 claims description 20
- 238000012549 training Methods 0.000 claims description 16
- 108010001267 Protein Subunits Proteins 0.000 claims description 5
- 238000000605 extraction Methods 0.000 claims description 4
- 238000012545 processing Methods 0.000 abstract description 10
- 230000000875 corresponding effect Effects 0.000 description 62
- 238000012706 support-vector machine Methods 0.000 description 24
- 238000010586 diagram Methods 0.000 description 11
- 238000004590 computer program Methods 0.000 description 8
- 206010020400 Hostility Diseases 0.000 description 5
- 238000012986 modification Methods 0.000 description 5
- 230000004048 modification Effects 0.000 description 5
- 238000002790 cross-validation Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 239000000284 extract Substances 0.000 description 3
- 238000012417 linear regression Methods 0.000 description 3
- 206010002368 Anger Diseases 0.000 description 2
- 241000406668 Loxodonta cyclotis Species 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 235000013399 edible fruits Nutrition 0.000 description 2
- 238000007477 logistic regression Methods 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 230000007935 neutral effect Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 206010029216 Nervousness Diseases 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
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
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/48—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use
- G10L25/51—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination
- G10L25/63—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination for estimating an emotional state
-
- 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
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
-
- 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
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/27—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique
-
- 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
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/27—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique
- G10L25/30—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique using neural networks
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Signal Processing (AREA)
- Multimedia (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Child & Adolescent Psychology (AREA)
- General Health & Medical Sciences (AREA)
- Hospice & Palliative Care (AREA)
- Psychiatry (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention discloses a kind of personality recognition methods and device, the method, including：Obtain the sound bite of measurand；The acoustic feature information in each voice sub-piece is extracted respectively, wherein the voice sub-piece is divided to obtain according to the sound bite；For each voice sub-piece, the acoustic feature information in the voice sub-piece is handled using preset algorithm, determines the corresponding type of emotion of voice sub-piece；According to the corresponding type of emotion of each voice sub-piece determined, the personality recognition result of measurand is determined.Using method provided by the invention, point processing such as carry out by the sound bite to measurand, and for each voice sub-piece obtained after decile, acoustic feature information is provided, then the acoustic feature information is handled, not only accurately the personality of measurand is analyzed, moreover it is possible to obtain the personality recognition result of measurand in time.
Description
Technical field
The present invention relates to personality analysis technical field more particularly to a kind of personality recognition methods and devices.
Background technology
In recent years, psychological educational circles is formd on personality descriptive model than more consistent common recognition, it is proposed that personality it is big by five
Model.Five speciality of big five model are respectively：Nervousness, extropism, opening, compatibility and doing one's duty property, wherein Mei Gete
Matter contains six sub- dimensions.Indignation and hostility are to weigh unstrung sub- dimension respectively, reaction be personal experience's indignation and
The tendency of correlated condition, such as setback and pain, measurement is that a people generates angry easy degree.Indignation and hostility have height
It is intrinsic and low intrinsic.The relatively high people of score shows as being easy to be under fire, especially when experiencing the treatment that oneself is subject to not
Animosity can be full of after just, the relatively high people of score can become irascible, indignation and feel to baffle.The relatively low crowd of score,
It then can easily control the emotion of oneself, it is not easy to it is under fire and angry, in daily life, the relatively low crowd's meeting of score
Close friend is showed, has an amiable disposition, be not easy the image flared up.Generally speaking, unstrung sub- dimension can reflect the choler of people,
Indignation and the high people of hostility score show as irritability, and indignation and the relatively low people of hostility score show as not irritability.
The method for measuring big five models, five speciality in the prior art is scaling method, and with indignation and hostility, this sub- dimension is
Example illustrates, that is, is directed to this sub- dimension, measurand needs to complete this problem corresponding to sub- dimension, then by measuring people
The score of this sub- dimension of member's statistics, obtains the characteristic of this sub- dimension, so as to show whether measurand is with easy
Anger personality.In addition, for the accurate anlage for assessing measurand and showing, when to the answer of each problem
Using ranking mode, and the range evaluated is generally divided into five grades, i.e., from very different meaning to agreeing to very much.Existing skill
Common scale has NEO-PI-R etc. in the scaling method used in art.
However when being measured in the prior art to five features of measurand using scaling method, can generally exist following
Problem：(1) in person to person's interactive process, the choler timeliness that measurand is obtained by way of filling in scale is relatively low,
It can not timely feedback whether measurand has choler personality；(2) it is only filled in the people of measurand Long Term Contact
Whether the accurate reaction measurand of answer ability of the problems in scale has choler personality.
How accurately and in time in conclusion, whether reflection measurand has the technology that choler personality is urgently to be resolved hurrily
One of problem.
Invention content
A kind of personality recognition methods of offer of the embodiment of the present invention and device, accurately to analyze the personality of measurand, and
Improve the timeliness of character analysis.
The embodiment of the present invention provides a kind of personality recognition methods, including：
Obtain the sound bite of measurand；
The acoustic feature information in each voice sub-piece is extracted respectively, wherein the voice sub-piece is according to institute's predicate
Tablet section is divided；
For each voice sub-piece, using preset algorithm to the acoustic feature information in the voice sub-piece at
Reason, determines the corresponding type of emotion of voice sub-piece；
According to the corresponding type of emotion of each voice sub-piece determined, the mood personality identification knot of measurand is determined
Fruit.
Preferably, each voice sub-piece is directed to, using preset algorithm to the acoustic feature information in the voice sub-piece
It is handled, determines the corresponding character type of voice sub-piece, specifically include：
The acoustic feature information is handled using preset algorithm, determines the corresponding mood of the acoustic feature information
Measured value；
Compare the emotional measurement value and obtains comparison result with default mood threshold value；
According to comparison result, the corresponding type of emotion of voice sub-piece is determined.
Preferably, being handled the acoustic feature information using preset algorithm, the acoustic feature information pair is determined
The emotional measurement value answered, specifically includes：
The acoustic feature information is handled according to following formula, determines the corresponding mood of the acoustic feature information
Measured value：
Wherein, x indicates the acoustic feature information；
F (x) indicates the corresponding emotional measurement value of the acoustic feature information；
Indicate optimal Lagrange multiplier vector；
b*Indicate optimal hyperlane intercept；
yiFor predetermined value, value is { -1 ,+1 }；
N indicates the quantity of sample in training set；
K (x, z) is gaussian kernel function, andWherein z indicates the acoustic feature information
Mean value；σ indicates the standard deviation of the acoustic feature information.
Further, it according to comparison result, determines the corresponding type of emotion of voice sub-piece, specifically includes：
If the comparison result, which is the emotional measurement value, is more than or equal to the default mood threshold value, it is determined that the voice
The corresponding type of emotion of sub-piece is negative-morality；Or
If the comparison result, which is the emotional measurement value, is less than the default mood threshold value, it is determined that the voice sub-pieces
The corresponding type of emotion of section is positive mood.
Further, according to the corresponding type of emotion of each voice sub-piece determined, the personality of measurand is determined
Recognition result specifically includes：
According to the corresponding type of emotion of each voice sub-piece determined, voice that type of emotion is negative-morality is determined
The quantity of segment；
Judge whether the type of emotion is more than preset quantity threshold value for the quantity of the voice sub-piece of negative-morality；
If it is, determining that the personality recognition result of measurand is irritability personality；
If it is not, then determining the personality recognition result of measurand for not irritability personality.
The embodiment of the present invention provides a kind of personality identification device, including：
Acquiring unit, the sound bite for obtaining measurand；
Extraction unit, for extracting the acoustic feature information in each voice sub-piece respectively, wherein the voice sub-pieces
Section is divided to obtain according to the sound bite；
First determination unit, for being directed to each voice sub-piece, using preset algorithm to the sound in the voice sub-piece
It learns characteristic information to be handled, determines the corresponding type of emotion of voice sub-piece；
Second determination unit, for according to the corresponding type of emotion of each voice sub-piece determined, determining tested pair
The personality recognition result of elephant.
Preferably, first determination unit, specifically includes：First determination subelement, comparing subunit and second determine
Subelement, wherein：
First determination subelement determines institute for being handled the acoustic feature information using preset algorithm
State the corresponding emotional measurement value of acoustic feature information；
The comparing subunit obtains comparison result for the emotional measurement value and default mood threshold value；
Second determination subelement, for according to comparison result, determining the corresponding type of emotion of voice sub-piece.
Preferably, first determination subelement, is specifically used for carrying out the acoustic feature information according to following formula
Processing, determines the corresponding emotional measurement value of the acoustic feature information：
Wherein, x indicates the acoustic feature information；
F (x) indicates the corresponding emotional measurement value of the acoustic feature information；
Indicate optimal Lagrange multiplier vector；
b*Indicate optimal hyperlane intercept；
yiFor predetermined value, value is { -1 ,+1 }；
N indicates the quantity of sample in training set；
K (x, z) is gaussian kernel function, andWherein z indicates the acoustic feature information
Mean value；σ indicates the standard deviation of the acoustic feature.
Further, second determination subelement, if being the emotional measurement value specifically for the comparison result
More than or equal to the default mood threshold value, it is determined that the corresponding type of emotion of voice sub-piece is negative-morality；Or if institute
It is that the emotional measurement value is less than the default mood threshold value to state comparison result, it is determined that the corresponding mood class of the voice sub-piece
Type is positive mood.
Further, second determination unit, specifically includes：Third determination subelement, judgment sub-unit, the 4th determine
Subelement and the 5th determination subelement, wherein：
The third determination subelement, for according to the corresponding type of emotion of each voice sub-piece determined, determining
Type of emotion is the quantity of the voice sub-piece of negative-morality；
The judgment sub-unit, for judging whether the type of emotion is more than for the quantity of the voice sub-piece of negative-morality
Preset quantity threshold value；
4th determination subelement, if the judging result for the judgment sub-unit is yes, it is determined that measurand
Personality recognition result is irritability personality；
5th determination subelement, if the judging result for the judgment sub-unit is no, it is determined that measurand
Personality recognition result is not irritability personality.
Personality recognition methods provided in an embodiment of the present invention and device, after the sound bite for obtaining measurand, to institute
It states sound bite to be divided, is then directed to each voice sub-piece obtained after dividing, extracts the sound in the voice sub-piece
Characteristic information is learned, according to the acoustic feature information, the acoustic feature information is handled using preset algorithm, determines voice
The corresponding type of emotion of segment determines the property of measurand after the completion of waiting for that the corresponding type of emotion of each voice sub-piece determines
Lattice recognition result can not only accurately analyze the personality of measurand, moreover it is possible to obtain the personality identification of measurand in time
As a result, more improving user experience.
Other features and advantages of the present invention will be illustrated in the following description, also, partly becomes from specification
It obtains it is clear that understand through the implementation of the invention.The purpose of the present invention and other advantages can be by the explanations write
Specifically noted structure is realized and is obtained in book, claims and attached drawing.
Description of the drawings
Attached drawing described herein is used to provide further understanding of the present invention, and constitutes the part of the present invention, this hair
Bright illustrative embodiments and their description are not constituted improper limitations of the present invention for explaining the present invention.In the accompanying drawings：
Fig. 1 a are the flow diagram for the personality recognition methods that the embodiment of the present invention one provides；
Fig. 1 b are the flow signal for the method that the corresponding type of emotion of voice sub-piece is determined in the embodiment of the present invention one
Figure；
Fig. 1 c are the flow diagram of the method for the personality recognition result that measurand is determined in the embodiment of the present invention one；
Fig. 2 is the structural schematic diagram of personality identification device provided by Embodiment 2 of the present invention.
Specific implementation mode
A kind of personality recognition methods of offer of the embodiment of the present invention and device, accurately to analyze the personality of measurand, together
The timeliness that Shi Tigao analyzes measurand personality.
In invention, it is to be understood that in involved term：
1, emotional measurement value：Emotional measurement value of the present invention is for reflecting whether measurand has irritability personality
Evaluation parameter.
2, type of emotion：Type of emotion of the present invention is to reflect whether measurand has irritability personality, institute
It states character type and is divided into positive mood and negative-morality.
Below in conjunction with Figure of description, preferred embodiment of the present invention will be described, it should be understood that described herein
Preferred embodiment only for the purpose of illustrating and explaining the present invention and is not intended to limit the present invention, and in the absence of conflict, this hair
The feature in embodiment and embodiment in bright can be combined with each other.
Embodiment one
As shown in Figure 1a, be the flow diagram of personality recognition methods that the embodiment of the present invention one provides, may include with
Lower step：
S11, the sound bite for obtaining measurand.
S12, acoustic feature information in each voice sub-piece is extracted respectively, wherein the voice sub-piece is according to institute
State what sound bite was divided.
When it is implemented, in order to improve the accuracy of personality recognition result, after obtaining sound bite, by the voice sheet
Section is divided into the voice sub-piece of identical duration, for example, when a length of 10 seconds sound bites can be with when point operation such as carrying out
It is halved, trisection can also be carried out, specifically determined by user experience.
Preferably, when being divided to the sound bite, when point operation such as can carry out to the sound bite.Tool
Body, it by preset duration etc. point interception sound bite and can be cut after receiving the sound bite of measurand
Voice sub-piece after taking, same section is not present in the voice sub-piece after the interception, or when intercepting sound bite, presses
The voice sub-piece obtained according to preset duration etc. point interception may exist same section, it is possible thereby to preferably show that personality identifies
As a result.For convenience, the embodiment of the present invention is illustrated for point processing such as carrying out to sound bite.
Preferably, the duration of the voice sub-piece after decile can be 2 seconds.I.e.：If a length of 20 when the sound bite
Second, then it can carry out ten decile operations.
When it is implemented, it is directed to each voice sub-piece, it, can be with when extracting the acoustic feature information of the voice sub-piece
Acoustic feature information is extracted using openSMILE Open-Source Tools.Specifically, the acoustic feature information can be based on 384 dimensions
Acoustic feature information.16 Wiki eigens are extracted first with openSMILE, specifically include zero-crossing rate, energy root mean square, base
Frequently, harmonic to noise ratio and 12 Jan Vermeer frequency cepstral coefficients and its first-order difference and statistical property, 1 institute of particular content reference table
Show.
Preferably, when extracting essential characteristic, can be extracted based on frame level, the essential characteristic and its a scale
Divide through 12 statistical functions, finally obtains 16*2*12=384 dimensional features.
Table 1
Preferably, the acoustic feature information is included at least with the next item down：Zero-crossing rate, energy root mean square, fundamental frequency, harmonic wave are made an uproar
Acoustic ratio and 12 Jan Vermeer frequency cepstral coefficients.
Specifically, bilingual relationship is in table 1：LLD(low-level descriptors)：Bottom describes the factor,
Functional：Function, Mean：Mean value, Standard devision：Standard deviation, Kurtosis：Kurtosis, Skewness：Partially
Gradient, Extremes：Value, rel.position, range：Maximum value minimum, the position of maximum frame, minimum frame
Position, maxima and minima range, Linear regression:offset,slope,MSE：The linear regression of biasing,
The linear regression of slope, mean square error linear return.
S13, it is directed to each voice sub-piece, the acoustic feature information in the voice sub-piece is carried out using preset algorithm
Processing, determines the corresponding type of emotion of voice sub-piece.
It, can be with when it is implemented, the corresponding type of emotion of voice sub-piece can be determined according to method shown in Fig. 1 b
Include the following steps：
S131, the acoustic feature information is handled using preset algorithm, determines that the acoustic feature information corresponds to
Emotional measurement value.
When it is implemented, the embodiment of the present invention uses SVM (the Support Vector with gaussian kernel function
Machine, support vector machines) algorithm handles the acoustic feature information, determine that the acoustic feature information is corresponding
Emotional measurement value.
It should be noted that the basic principle of SVM algorithm is：Classification prediction is carried out to data based on Statistical Learning Theory.
It is dedicated to finding the generalization ability that structuring least risk further increases learning machine, to reach empiric risk and fiducial range
Minimum, it is final so that in the case where statistical sample amount is fewer, can also obtain good learning effect.The present invention is real
The SVM algorithm that example uses is applied as nonlinear support vector machines.I.e.：Kernel function is applied in SVM algorithm, uses one first
The data for becoming linearly inseparable of changing commanders are mapped to a new higher dimensional space from original space data are become linear separability, then
The rule of the method for linear classification study training data is used in new higher dimensional space.Kernel function is applied in support vector machines
(SVM) in, exactly a new feature is mapped to using a nonlinear original input space-theorem in Euclid space Rn that changes commanders that becomes
Space-Hilbert space H can be thus that the hypersurface inside original space becomes in new Hilbert space
Hyperplane.
Specifically, gaussian kernel function provided in an embodiment of the present invention is one kind in kernel function.Certainly it can also use
Polynomial kernel function applies it in SVM algorithm and executes personality recognition methods provided by the invention.
Emotional measurement value described in S132, comparison obtains comparison result with default mood threshold value.
S133, according to comparison result, determine the corresponding type of emotion of voice sub-piece.
Preferably, when executing step S133, if the comparison result is the emotional measurement value be more than or equal to it is described
Default mood threshold value, it is determined that the corresponding type of emotion of voice sub-piece is negative-morality；Or
If the comparison result, which is the emotional measurement value, is less than the default mood threshold value, it is determined that the voice sub-pieces
The corresponding type of emotion of section is positive mood.
Preferably, the corresponding type of emotion of voice sub-piece can also be determined using following methods：If the comparison
As a result it is that the emotional measurement value is equal to the prediction mood threshold value, it is determined that the corresponding type of emotion of voice sub-piece is negative
Mood；Or if the comparison result, which is the emotional measurement value, is not equal to the default mood threshold value, it is determined that the voice
The corresponding type of emotion of sub-piece is positive mood.Specifically, the default mood threshold value can be -1.
Specifically, the positive mood can be, but not limited to include neutral (neutral) and joy (happy)；It is described negative
Mood can be, but not limited to include hot anger (angry) and cold anger (cold violence).
For a better understanding of the present invention, the embodiment of the present invention one with preset algorithm be the support with gaussian kernel function to
It is illustrated for amount machine SVM algorithm, it, can be true according to formula (1) for each voice sub-piece when executing step S131
Determine the corresponding emotional measurement value of the acoustic feature information：
Wherein, N indicates the quantity of the acoustic feature,Indicate optimal Lagrange multiplier vector, b*Indicate optimal super
Plane intercept, yiFor setting value, value is { -1 ,+1 }.
In formula, K (x, z) is gaussian kernel function, shown in expression formula reference formula (2)：
X indicates the acoustic feature information in formula (1)；F (x) indicates the corresponding emotional measurement of the acoustic feature information
Value, z indicate the mean value of the acoustic feature information；σ indicates the standard deviation of the acoustic feature information.
It, will be in the acoustic feature information input to the formula (1) based on above-mentioned formula (1), you can determine described
The corresponding emotional measurement value of acoustic feature information.
When it is implemented, before executing personality recognition methods provided by the invention, it is also necessary to being imputed in advance in the present invention
The relevant parameter that method is related to is determined, and substantially process is：Sound bite (training set data) is obtained first from corpus,
After obtaining sound bite, the sound bite is subjected to decile, for the voice sub-piece after each decile, determines voice
Then the corresponding type of emotion of segment determines the personality recognition result of measurand, i.e., according to type of emotion：According to known tested
The personality recognition result of object can determine measurand current speech sub-piece after obtaining the voice sub-piece of measurand
Type of emotion, i.e.,：It is input in preset algorithm with the voice sub-piece of known type of emotion in advance, based on this determination preset algorithm
Middle relevant parameter.
For a better understanding of the present invention, it is illustrated by taking the SVM algorithm with gaussian kernel function as an example, is executing this hair
Before bright, it is thus necessary to determine that a in above-mentioned formula (1)i *, z, σ and b*Occurrence.It, can be by based on sound bite in corpus
The phonetic decision such as is subjected at point operation according to the method for step S12 of the present invention, can determine that the voice sub-piece corresponds in advance
Type of emotion.Then it is directed to each voice sub-piece, extracts the acoustic feature information in the voice sub-piece, you can to obtain
X in formula (1).It is unknown due to there is 4 parameters in formula (1), it is therefore desirable to utilize at least four sound bite and its right
The type of emotion answered determines the parameter in formula (1), for example, if type of emotion is positive mood, the f (x) in formula (1)
It is+1, if type of emotion is negative-morality, the f (x) in formula (1) is -1.So far, it may be determined that go out with gaussian kernel function
SVM algorithm in relevant parameter.
Further, when determining the relevant parameter of SVM algorithm according to the voice sub-piece of known type of emotion, in advance will
The voice sub-piece of input is labeled, and NEO-PI-R methods mark voice sub-piece specifically may be used, in order to ensure personality
The corresponding type of emotion of voice sub-piece of the accuracy of recognition result, selection should not be single.
Preferably, in the relevant parameter in determining formula (1), it can also be using the method for crossing over many times verification come really
It is fixed, for example, the number of cross validation can be, but not limited to be 5 times.It is possible thereby to avoid the randomness of parameter determined.
When it is implemented, being illustrated by taking 5 cross validations as an example, 5 voice sub-pieces are obtained first from corpus,
The corresponding type of emotion of sound bite is determined respectively, is then directed to the acoustics spy that each voice sub-piece extracts the voice sub-piece
Reference ceases, and composing training collection, the training set includes training sample, and the quantity for the sample for being included is N.The trained sample
This is the corresponding acoustic feature information of voice sub-piece and type of emotion.Based on the basic principle of 5 cross validations, from training set
In 4 groups of training samples of any selection as training data, another group is used as test data, carries out a cross validation and obtains one group
Then the parameter value of SVM algorithm repeats 4 times, respectively obtains the parameter value of 4 groups of SVM algorithms, according to according to 5 groups of obtained SVM
The parameter value of algorithm determines parameter value of the optimal class value as final SVM algorithm again.
Preferably, in order to improve the accuracy of personality recognition result, in SVM algorithm provided in an embodiment of the present invention, may be used also
To introduce penalty coefficient C, then the parameter value in SVM algorithm is determined according to formula (3)：
Wherein N indicates training set total sample number amount；aiIt indicates,：Lagrange multiplier vector, hi,hjFor sample in training set
The corresponding value of corresponding personality recognition result, value be,：{ -1 ,+1 }, shown in the constraints reference formula (4) of formula (3)：
When it is implemented, if the corresponding personality recognition result of sample is irritability personality, corresponding value in training set
It is -1；Otherwise its corresponding value is+1.When determining penalty coefficient C, grid data service may be used and select best punishment system
Then number determines optimal Lagrange multiplier vector in SVM algorithm further according to formula (3) and (4)Then it chooses to be located at and open
Section (0, C)Calculate optimal hyperlane interceptBased on this, by introduce penalty coefficient,
So that when executing personality recognition methods provided in an embodiment of the present invention, the SVM algorithm that can be determined according to penalty coefficient obtains
More accurate personality recognition result is obtained, user experience is improved.
Preferably, in order to improve the accuracy of personality recognition result, in SVM algorithm provided in an embodiment of the present invention, may be used also
To introduce kernel function gamma parameter γ, i.e., the Gaussian kernel in the SVM algorithm used in the embodiment of the present invention with gaussian kernel function
Function can introduce gamma parameter γ, that is, the gaussian kernel function expression formula for introducing gamma parameters is：
When it is implemented, gamma parameter γ can be determined using grid data service, based on this using formula (5) as new
Gaussian kernel function substitute into formula (1), then determine the type of emotion of the voice sub-piece, it is to be determined go out each voice sub-piece
After corresponding type of emotion, the personality recognition result of measurand is further determined that so that according to the Gauss with gamma parameters
The SVM algorithm of kernel function determines more accurate personality recognition result, improves user experience.
Preferably, the corresponding emotional measurement value of the acoustic feature information can also be determined using following preset algorithm：ELM
(Extreme Learning Machine, extreme learning machine) algorithm, GMM (Gaussian Mixture Model, Gaussian Mixture
Model), ANN (Artificial Neural Machine, artificial neural network) algorithms and LR (Logistic
Regression, logistic regression) algorithm.
The corresponding type of emotion of each voice sub-piece that S14, basis are determined determines the personality identification knot of measurand
Fruit.
Preferably, can be according to method shown in Fig. 1 c according to the corresponding mood class of each voice sub-piece determined
Type determines the personality recognition result of measurand, includes the following steps：
The corresponding type of emotion of each voice sub-piece that S141, basis are determined determines that type of emotion is negative-morality
The quantity of voice sub-piece.
S142, judge whether the type of emotion is more than preset quantity threshold value for the quantity of the voice sub-piece of negative-morality,
If so, thening follow the steps S143；It is no to then follow the steps S144.
S143, determine that the personality recognition result of measurand is irritability personality.
S144, determine the personality recognition result of measurand for not irritability personality.
When it is implemented, the quantity for working as the voice sub-piece for determining that the type of emotion is negative-morality is more than preset quantity
When threshold value, then assert that measurand has the characteristics that irritability, excited type personality easily show angry and indignation, temper ratio
It is more irascible；When determine the type of emotion be negative-morality voice sub-piece quantity be less than or equal to preset quantity threshold value
When, then assert that measurand has the characteristics that not irritability, mood high order smooth pattern personality show as being not easy ignition and angry, temper ratio
It is more amiable.
The personality recognition methods that the embodiment of the present invention one provides, after the sound bite for obtaining measurand, by institute's predicate
Tablet section such as carries out at point operation, is then directed to each voice sub-piece after decile, and the acoustics extracted in the voice sub-piece is special
Reference ceases, and according to the acoustic feature information, handles the acoustic feature information using preset algorithm, determines the voice sub-piece
Corresponding type of emotion, after the completion of waiting for that the corresponding type of emotion of each voice sub-piece determines, if it is determined that go out the mood class
Type is that the quantity of the voice sub-piece of negative-morality is more than preset quantity threshold value, it is determined that the personality recognition result of measurand is easy
Anger personality；Otherwise determine that the personality recognition result of measurand for not irritability personality, can not only accurately analyze tested pair
The personality of elephant, moreover it is possible to which the personality recognition result for obtaining measurand in time more improves user experience.
Embodiment two
Based on same inventive concept, a kind of personality identification device is additionally provided in the embodiment of the present invention, due to above-mentioned apparatus
The principle solved the problems, such as is similar to personality recognition methods, therefore the implementation of above-mentioned apparatus may refer to the implementation of method, repetition
Place repeats no more.
As shown in Fig. 2, for the structural schematic diagram of personality identification device provided by Embodiment 2 of the present invention, including：It obtains single
Member 21, extraction unit 22, the first determination unit 23 and the second determination unit 24, wherein：
Acquiring unit 21, the sound bite for obtaining measurand；
Extraction unit 22, for extracting the acoustic feature information in each voice sub-piece respectively, wherein the voice is sub
Segment is divided to obtain according to the sound bite；
First determination unit 23, for being directed to each voice sub-piece, using preset algorithm in the voice sub-piece
Acoustic feature information is handled, and determines the corresponding type of emotion of voice sub-piece；
Second determination unit 24, for according to the corresponding type of emotion of each voice sub-piece determined, determining tested
The personality recognition result of object.
When it is implemented, first determination unit 23, specifically includes：First determination subelement, comparing subunit and
Two determination subelements, wherein：
First determination subelement determines institute for being handled the acoustic feature information using preset algorithm
State the corresponding emotional measurement value of acoustic feature information；
The comparing subunit obtains comparison result for the emotional measurement value and default mood threshold value；
Second determination subelement, for according to comparison result, determining the corresponding type of emotion of voice sub-piece.
Preferably, first determination subelement, is specifically used for carrying out the acoustic feature information according to following formula
Processing, determines the corresponding emotional measurement value of the acoustic feature information：
Wherein, x indicates the acoustic feature information；
F (x) indicates the corresponding emotional measurement value of the acoustic feature information；
Indicate optimal Lagrange multiplier vector；
b*Indicate optimal hyperlane intercept；
yiFor setting value, value is { -1 ,+1 }；
N indicates the quantity of sample in training set；
K (x, z) is gaussian kernel function, andWherein z indicates the acoustic feature information
Mean value；σ indicates the standard deviation of the acoustic feature information.
Preferably, second determination subelement, if being that the emotional measurement value is big specifically for the comparison result
In equal to the default mood threshold value, it is determined that the corresponding type of emotion of voice sub-piece is negative-morality；Or it is if described
Comparison result is that the emotional measurement value is less than the default mood threshold value, it is determined that the corresponding type of emotion of voice sub-piece
For positive mood.
Preferably, second determination unit 24, specifically includes：Third determination subelement, judgment sub-unit, the 4th determine
Subelement and the 5th determination subelement, wherein：
The third determination subelement, for according to the corresponding type of emotion of each voice sub-piece determined, determining
Type of emotion is the quantity of the voice sub-piece of negative-morality；
The judgment sub-unit, for judging whether the type of emotion is more than for the quantity of the voice sub-piece of negative-morality
Preset quantity threshold value；
4th determination subelement, if the judging result for the judgment sub-unit is yes, it is determined that measurand
Personality recognition result is irritability personality；
5th determination subelement, if the judging result for the judgment sub-unit is no, it is determined that measurand
Personality recognition result is not irritability personality.
Preferably, the acoustic feature information is included at least with the next item down：Zero-crossing rate, energy root mean square, fundamental frequency, harmonic wave are made an uproar
Acoustic ratio and 12 Jan Vermeer frequency cepstral coefficients.
For convenience of description, above each section is divided by function describes respectively for each module (or unit).Certainly, exist
Implement the function of each module (or unit) can be realized in same or multiple softwares or hardware when the present invention.
The personality identification device that embodiments herein is provided can be realized by computer program.Those skilled in the art
It should be appreciated that above-mentioned module dividing mode is only one kind in numerous module dividing modes, if being divided into other moulds
Block or non-division module all should be within the protection domains of the application as long as personality identification device has above-mentioned function.
It should be understood by those skilled in the art that, the embodiment of the present invention can be provided as method, system or computer program
Product.Therefore, complete hardware embodiment, complete software embodiment or reality combining software and hardware aspects can be used in the present invention
Apply the form of example.Moreover, the present invention can be used in one or more wherein include computer usable program code computer
The computer program production implemented in usable storage medium (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.)
The form of product.
The present invention be with reference to according to the method for the embodiment of the present invention, the flow of equipment (system) and computer program product
Figure and/or block diagram describe.It should be understood that can be realized by computer program instructions every first-class in flowchart and/or the block diagram
The combination of flow and/or box in journey and/or box and flowchart and/or the block diagram.These computer programs can be provided
Instruct the processor of all-purpose computer, special purpose computer, Embedded Processor or other programmable data processing devices to produce
A raw machine so that the instruction executed by computer or the processor of other programmable data processing devices is generated for real
The device for the function of being specified in present one flow of flow chart or one box of multiple flows and/or block diagram or multiple boxes.
These computer program instructions, which may also be stored in, can guide computer or other programmable data processing devices with spy
Determine in the computer-readable memory that mode works so that instruction generation stored in the computer readable memory includes referring to
Enable the manufacture of device, the command device realize in one flow of flow chart or multiple flows and/or one box of block diagram or
The function of being specified in multiple boxes.
These computer program instructions also can be loaded onto a computer or other programmable data processing device so that count
Series of operation steps are executed on calculation machine or other programmable devices to generate computer implemented processing, in computer or
The instruction executed on other programmable devices is provided for realizing in one flow of flow chart or multiple flows and/or block diagram one
The step of function of being specified in a box or multiple boxes.
Although preferred embodiments of the present invention have been described, it is created once a person skilled in the art knows basic
Property concept, then additional changes and modifications can be made to these embodiments.So it includes excellent that the following claims are intended to be interpreted as
It selects embodiment and falls into all change and modification of the scope of the invention.
Obviously, various changes and modifications can be made to the invention without departing from essence of the invention by those skilled in the art
God and range.In this way, if these modifications and changes of the present invention belongs to the range of the claims in the present invention and its equivalent technologies
Within, then the present invention is also intended to include these modifications and variations.
Claims (10)
1. a kind of personality recognition methods, which is characterized in that including：
Obtain the sound bite of measurand；
The acoustic feature information in each voice sub-piece is extracted respectively, wherein the voice sub-piece is according to the voice sheet
What Duan Jinhang was divided；
For each voice sub-piece, the acoustic feature information in the voice sub-piece is handled using preset algorithm, really
The fixed corresponding type of emotion of voice sub-piece；
According to the corresponding type of emotion of each voice sub-piece determined, the personality recognition result of measurand is determined.
2. the method as described in claim 1, which is characterized in that each voice sub-piece is directed to, using preset algorithm to the language
Acoustic feature information in phone segment is handled, and is determined the corresponding type of emotion of voice sub-piece, is specifically included：
The acoustic feature information is handled using preset algorithm, determines the corresponding emotional measurement of the acoustic feature information
Value；
Compare the emotional measurement value and obtains comparison result with default mood threshold value；
According to comparison result, the corresponding type of emotion of voice sub-piece is determined.
3. method as claimed in claim 2, which is characterized in that using preset algorithm to the acoustic feature information at
Reason, determines the corresponding emotional measurement value of the acoustic feature information, specifically includes：
The acoustic feature information is handled according to following formula, determines the corresponding emotional measurement of the acoustic feature information
Value：
Wherein, x indicates the acoustic feature information；
F (x) indicates the corresponding emotional measurement value of the acoustic feature information；
ai *Indicate optimal Lagrange multiplier vector；
b*Indicate optimal hyperlane intercept；
yiFor predetermined value, value is { -1 ,+1 }；
N indicates the quantity of sample in training set；
K (x, z) is gaussian kernel function, andWherein z indicates the mean value of the acoustic feature information；
σ indicates the standard deviation of the acoustic feature.
4. method as claimed in claim 2, which is characterized in that according to comparison result, determine the corresponding feelings of voice sub-piece
Thread type, specifically includes：
If the comparison result, which is the emotional measurement value, is more than or equal to the default mood threshold value, it is determined that the voice sub-pieces
The corresponding type of emotion of section is negative-morality；Or
If the comparison result, which is the emotional measurement value, is less than the default mood threshold value, it is determined that the voice sub-piece pair
The type of emotion answered is positive mood.
5. method as claimed in claim 4, which is characterized in that according to the corresponding mood class of each voice sub-piece determined
Type determines the personality recognition result of measurand, specifically includes：
According to the corresponding type of emotion of each voice sub-piece determined, determine that type of emotion is the voice sub-piece of negative-morality
Quantity；
Judge whether the type of emotion is more than preset quantity threshold value for the quantity of the voice sub-piece of negative-morality；
If it is, determining that the personality recognition result of measurand is irritability personality；
If it is not, then determining the personality recognition result of measurand for not irritability personality.
6. a kind of personality identification device, which is characterized in that including：
Acquiring unit, the sound bite for obtaining measurand；
Extraction unit, for extracting the acoustic feature information in each voice sub-piece respectively, wherein the voice sub-piece is
It is divided according to the sound bite；
First determination unit, it is special to the acoustics in the voice sub-piece using preset algorithm for being directed to each voice sub-piece
Reference breath is handled, and determines the corresponding type of emotion of voice sub-piece；
Second determination unit, for according to the corresponding type of emotion of each voice sub-piece determined, determining measurand
Personality recognition result.
7. device as claimed in claim 6, which is characterized in that first determination unit specifically includes：First determines that son is single
Member, comparing subunit and the second determination subelement, wherein：
First determination subelement determines the sound for being handled the acoustic feature information using preset algorithm
Learn the corresponding emotional measurement value of characteristic information；
The comparing subunit obtains comparison result for the emotional measurement value and default mood threshold value；
Second determination subelement, for according to comparison result, determining the corresponding type of emotion of voice sub-piece.
8. device as claimed in claim 7, which is characterized in that first determination subelement is specifically used for according to following public affairs
Formula handles the acoustic feature information, determines the corresponding emotional measurement value of the acoustic feature information：
Wherein, x indicates the acoustic feature information；
F (x) indicates the corresponding emotional measurement value of the acoustic feature information；
ai *Indicate optimal Lagrange multiplier vector；
b*Indicate optimal hyperlane intercept；
yiFor setting value, value is { -1 ,+1 }；
N indicates the quantity of sample in training set；
K (x, z) is gaussian kernel function, andWherein z indicates the mean value of the acoustic feature information；
σ indicates the standard deviation of the acoustic feature information.
9. device as claimed in claim 7, which is characterized in that
Second determination subelement, if specifically for the comparison result be the emotional measurement value be more than or equal to it is described pre-
If mood threshold value, it is determined that the corresponding type of emotion of voice sub-piece is negative-morality；Or if the comparison result is institute
It states emotional measurement value and is less than the default mood threshold value, it is determined that the corresponding type of emotion of voice sub-piece is positive mood.
10. device as claimed in claim 6, which is characterized in that second determination unit specifically includes：Third determines son
Unit, judgment sub-unit, the 4th determination subelement and the 5th determination subelement, wherein：
The third determination subelement, for according to the corresponding type of emotion of each voice sub-piece determined, determining mood
Type is the quantity of the voice sub-piece of negative-morality；
The judgment sub-unit is preset for judging the type of emotion for whether the quantity of the voice sub-piece of negative-morality is more than
Amount threshold；
4th determination subelement, if the judging result for the judgment sub-unit is yes, it is determined that the personality of measurand
Recognition result is irritability personality；
5th determination subelement, if the judging result for the judgment sub-unit is no, it is determined that the personality of measurand
Recognition result is not irritability personality.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710271654.1A CN108735232A (en) | 2017-04-24 | 2017-04-24 | A kind of personality recognition methods and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710271654.1A CN108735232A (en) | 2017-04-24 | 2017-04-24 | A kind of personality recognition methods and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108735232A true CN108735232A (en) | 2018-11-02 |
Family
ID=63934087
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710271654.1A Pending CN108735232A (en) | 2017-04-24 | 2017-04-24 | A kind of personality recognition methods and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108735232A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109451188A (en) * | 2018-11-29 | 2019-03-08 | 平安科技（深圳）有限公司 | Method, apparatus, computer equipment and the storage medium of the self-service response of otherness |
| WO2021027117A1 (en) * | 2019-08-15 | 2021-02-18 | 平安科技（深圳）有限公司 | Speech emotion recognition method and appartus, and computer-readable storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20080086791A (en) * | 2007-03-23 | 2008-09-26 | 엘지전자 주식회사 | Feeling recognition system based on voice |
| CN102623009A (en) * | 2012-03-02 | 2012-08-01 | 安徽科大讯飞信息技术股份有限公司 | Abnormal emotion automatic detection and extraction method and system on basis of short-time analysis |
| CN102831891A (en) * | 2011-06-13 | 2012-12-19 | 富士通株式会社 | Processing method and system for voice data |
| JP2015099304A (en) * | 2013-11-20 | 2015-05-28 | 日本電信電話株式会社 | Sympathy/antipathy location detecting apparatus, sympathy/antipathy location detecting method, and program |
| CN106250855A (en) * | 2016-08-02 | 2016-12-21 | 南京邮电大学 | A kind of multi-modal emotion identification method based on Multiple Kernel Learning |
-
2017
- 2017-04-24 CN CN201710271654.1A patent/CN108735232A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20080086791A (en) * | 2007-03-23 | 2008-09-26 | 엘지전자 주식회사 | Feeling recognition system based on voice |
| CN102831891A (en) * | 2011-06-13 | 2012-12-19 | 富士通株式会社 | Processing method and system for voice data |
| CN102623009A (en) * | 2012-03-02 | 2012-08-01 | 安徽科大讯飞信息技术股份有限公司 | Abnormal emotion automatic detection and extraction method and system on basis of short-time analysis |
| JP2015099304A (en) * | 2013-11-20 | 2015-05-28 | 日本電信電話株式会社 | Sympathy/antipathy location detecting apparatus, sympathy/antipathy location detecting method, and program |
| CN106250855A (en) * | 2016-08-02 | 2016-12-21 | 南京邮电大学 | A kind of multi-modal emotion identification method based on Multiple Kernel Learning |
Non-Patent Citations (1)
| Title |
|---|
| 张向君: "《信号分析与数据统计学习》", 28 February 2009, 哈尔滨工程大学出版社 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109451188A (en) * | 2018-11-29 | 2019-03-08 | 平安科技（深圳）有限公司 | Method, apparatus, computer equipment and the storage medium of the self-service response of otherness |
| WO2021027117A1 (en) * | 2019-08-15 | 2021-02-18 | 平安科技（深圳）有限公司 | Speech emotion recognition method and appartus, and computer-readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108597492B (en) | Phoneme synthesizing method and device | |
| Tsanas et al. | Objective automatic assessment of rehabilitative speech treatment in Parkinson's disease | |
| Bone et al. | Robust unsupervised arousal rating: A rule-based framework withknowledge-inspired vocal features | |
| Frauenfelder et al. | Simulating the time course of spoken word recognition: An analysis of lexical competition in TRACE | |
| Cummins et al. | Generalized two-stage rank regression framework for depression score prediction from speech | |
| CN110147732A (en) | Refer to vein identification method, device, computer equipment and storage medium | |
| CN110033022A (en) | Processing method, device and the storage medium of text | |
| CN110085211A (en) | Speech recognition exchange method, device, computer equipment and storage medium | |
| US11763174B2 (en) | Learning material recommendation method, learning material recommendation device, and learning material recommendation program | |
| Kurz-Milcke et al. | Heuristic decision making | |
| CN111382573A (en) | Method, apparatus, device and storage medium for answer quality assessment | |
| Fontes et al. | Classification system of pathological voices using correntropy | |
| WO2010021723A1 (en) | Content and quality assessment method and apparatus for quality searching | |
| US20230252315A1 (en) | Adjusting mental state to improve task performance | |
| CN109685104B (en) | Determination method and device for recognition model | |
| JP2018045062A (en) | Program, device and method automatically grading from dictation voice of learner | |
| Dsouza et al. | Chat with bots intelligently: A critical review & analysis | |
| Velmurugan et al. | Developing a fidelity evaluation approach for interpretable machine learning | |
| Shibata et al. | Analytic automated essay scoring based on deep neural networks integrating multidimensional item response theory | |
| CN108735232A (en) | A kind of personality recognition methods and device | |
| KR101745874B1 (en) | System and method for a learning course automatic generation | |
| Sundarprasad | Speech emotion detection using machine learning techniques | |
| Loughran et al. | Feature selection for speaker verification using genetic programming | |
| Banerjee et al. | Relation extraction using multi-encoder lstm network on a distant supervised dataset | |
| CN113705792A (en) | Personalized recommendation method, device, equipment and medium based on deep learning model |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181102 |