CN105212949A - A kind of method using skin pricktest signal to carry out culture experience emotion recognition - Google Patents
A kind of method using skin pricktest signal to carry out culture experience emotion recognition Download PDFInfo
- Publication number
- CN105212949A CN105212949A CN201510528090.6A CN201510528090A CN105212949A CN 105212949 A CN105212949 A CN 105212949A CN 201510528090 A CN201510528090 A CN 201510528090A CN 105212949 A CN105212949 A CN 105212949A
- Authority
- CN
- China
- Prior art keywords
- emotion
- signal
- sample
- skin pricktest
- emotion recognition
- 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
Landscapes
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention discloses a kind of method using skin pricktest signal to carry out culture experience emotion recognition, when carrying out emotional experience identification, gather the skin pricktest signal of experiencer, the KNN emotion recognition computation model trained before being inputted by signal, calculates the Euclidean distance between this sample and 175 samples gathered early stage.Choose 5 minimum Euclidean distances.According to the affective style of training sample corresponding to these 5 Euclidean distances, and the numerical result of these 5 Euclidean distances, judge the emotional experience classification of measured.The present invention is in the various parameter attributes of physiological signal, specify that 18 change with emotion the physiological feature parameter having corresponding relation, thus only need when carrying out emotion recognition at every turn to calculate a small amount of parameter attribute, greatly reduce the complexity that parameter calculates, improve computational efficiency.
Description
Technical field
The present invention relates to a kind of human emotion and know method for distinguishing.
Background technology
Emotion recognition is the key technology realizing harmonious man-machine interaction, its objective is the ability of giving computer recognizing user feeling.Research from society and cognitive psychology shows that, under relevant environmental stimuli, emotion can even unconsciously be aroused rapidly, easily, automatically.Affection computation is proposed in 1997 by the Picard professor of Massachusetts Institute Technology at first.The target of affection computation is the perception of imparting computer, understands and the ability showed emotion, thus more initiatively, friendlyly, excellent in voice and affection exchanges with people.Subsequently, affection computation causes rapidly the interest of artificial intelligence and computer realm expert, and becomes brand-new, a full of hope research field in recent years.The proposition of affection computation with develop rapidly, be the requirement due to man-machine interaction consonance on the one hand, wish that computer not only possesses the ability of listening, saying, seeing, read as people, and can understand and express the emotions such as pleasure, anger, sorrow, happiness; Also being the psychology based on strong calculating doctrine on the other hand, wishing calculating the inward world extending the pure man.After affection computation proposes, the emotion recognition based on facial expression, voice, posture and physiological signal is extensively being studied.Tomkins points out that facial exercises plays key player in emotional experience.But because facial exercises is subject to the impact that subjectivity covers up, so Picard thinks, based on the emotion recognition of physiological signal closer to the inherent psychological feelings of emotion.The team of Ekman during the emotion under controlled state (indignation, frightened, sad, glad, surprised and detest) have recorded tested heart rate, left and right finger temperature, skin resistance, the bent muscular tension of forearm, and sends the documents on the Science of nineteen eighty-three at first and set forth the evidence of discrete emotion ga s safety degree.Recent domestic researcheres are from skin conductivity (GSR), heart rate (HR), electrocardio (ECG), myoelectricity (EMG), breathe, skin temperature, the physiological signals such as pulse (BVP) carry out the research of affective state identification, demonstrate physiological signal and carry out the feasibility of feature identification emotion and establish the emotion recognition system with better recognition performance.Galvanic skin response (GSR) be the sympathetic activation change of reflection people the most effectively, the most responsive physical signs, be most extensive use the earliest in the world and obtain generally acceptedly leading psychological test index more.GSR have recorded the change of the skin conductivity of human body, reflects the change procedure of the sympathetic nervous system of human body, the change that can cause human body skin electricity to orthosympathetic stimulation produced by feelings induce.
Culture experience be people produce under the impact of Cultural Elements a kind of in activity, such as, see the scene of the Water-sprinkling Festival of the Dai nationality, people can produce a kind of mood of happiness of happiness; When watching Massacre in Nanjing memorial museum, people can produce a kind of angry emotion based on anti-Japanese feelings; When visit history commemorates museum, people perhaps can feel sorrowful to the harrowing experience of certain general; Come into a secret cavern, unknown, gloomy atmosphere can allow people produce a kind of emotion of fear.Emotion recognition is used for culture experience, identify people produce when culture is experienced in emotion be in affection computation another innovate.
Summary of the invention
Content of the present invention is to provide. a kind of method using skin pricktest signal to carry out culture experience emotion recognition.
In order to obtain above-mentioned purpose, by the following technical solutions:
. a kind of method using skin pricktest signal to carry out culture experience emotion recognition, " culture is experienced " in the present patent application refers to user's subjective experience for Related product when watching the cultural product such as picture, video, audio frequency with culture background, it is characterized in that:
Described algorithm comprises the steps:
S1: obtain happiness, sadness, indignation, fear, tranquil 5 class emotion samples that culture is experienced by experiment, every class emotion 35 samples, totally 175 samples;
S2: to the smoothing noise suppression preprocessing of skin pricktest signal gathered;
S3: correspondence 18 numerical characteristics extracting 175 samples respectively;
Wherein 18 numerical characteristics are as shown in the table:
S4: according to the eigenvalue of each sample, constructs the k nearest neighbor model of cognition for emotional experience identification.
The construction method of k nearest neighbor model of cognition is:
S1: adopt skin pricktest signal pickup assembly to gather user's skin pricktest signal, wherein sample rate is higher than 50HZ;
S2: the Euclidean distance of trying to achieve test sample book and 175 training samples; Euclidean distance computational methods between A sample and B sample are, each sample extraction 18 features; Then the eigenvalue of A sample is p1, p2, p3 respectively ... p18;
The eigenvalue of B sample is q1, q2, q3 respectively ... q18; So, the Euclidean distance d between A sample and B sample is
S3: in 175 Euclidean distances obtained, chooses 5 minimum Euclidean distances;
S4: the type of emotion of adding up training sample corresponding to these 5 Euclidean distances;
S4: calculate the ownership probability of tested sample for all kinds of emotion respectively;
Its computing formula is:
wherein θ
ifor current emotion is for the ownership probability of i-th kind of emotion; { 1,2,3,4,5} is corresponding glad, sad, indignation, frightened, tranquil five kinds of affective styles respectively for i ∈; K is the sample size belonging to the i-th type in 5 training samples nearest with test sample book; S is corresponding Euclidean distance numerical value.
Such as, in 5 training samples that the shortest Euclidean distance corresponding nearest with tested sample, there is the happiness sample that three are culture and experience, the sad sample that a culture is experienced, the tranquil sample that a culture is experienced, 2,5,10 respectively with the Euclidean distance of three happiness training samples, 3 with the Euclidean distance of a sad training sample, 20 with the Euclidean distance of a tranquil sample, the weighted value then belonging to happy emoticon is 1/2+1/5+1/10=0.8, the weighted value belonging to sad mood is 1/3=0.33, and the weighted value belonging to tranquil emotion is 1/20=0.5.Due to 0.8>0.5>0.33, so this test sample book belongs to happy emoticon sample.
Major advantage of the present invention comprises:
1, in the various parameter attributes of physiological signal, specify that 18 change with emotion the physiological feature parameter having corresponding relation, thus only need when carrying out emotion recognition at every turn to calculate a small amount of parameter attribute, greatly reduce the complexity that parameter calculates, improve computational efficiency.All not by clear and definite proposition in relevant parameter research in the past.
2, the method setting up emotion recognition model can rebuild for different user groups and different feelings induce stimulus materials, and construction method is simple, and data acquisition amount is relatively little.This makes the method can be transplanted to fast in the emotion recognition process of different scene and object.
3, use the emotion recognition model set up to carry out the accuracy of emotion recognition high, recognition speed is fast, can meet the computation complexity requirement of Real time identification, thus provides effective method for the real-time emotion recognition of Consumer's Experience.
Figure of description
Fig. 1 is the skin pricktest signal schematic representation before denoising;
Fig. 2 is the skin pricktest signal schematic representation after denoising;
Fig. 3 is k nearest neighbor method sample figure.
Detailed description of the invention
Below in conjunction with the drawings and specific embodiments, the present invention is further elaborated.
By different Numerical Methods, the numerical characteristics of a hundreds of different dimensions, different implication can be gone out skin pricktest signal extraction, and wherein only have the feature of minority and emotional experience to have potential relatedness.So far illustrate which signal characteristic has direct association with the identification of emotion physiology without any authoritative research conclusion.This research, by detection experiment, confirms to use 18 following skin pricktest signal combination, can be good at the situation of change reflecting emotion.This skin pricktest signal combination is:
Table 1: for the physiological feature list of emotion physiological signal identification
1. the feelings induce of emotion recognition model training data and skin pricktest signals collecting
Adopt the polygraph MP150 that Biopac company of the U.S. provides.This instrument gathers out GSR signal, and GSR useful signal frequency range mainly concentrates on below 0.2Hz, according to nyquist sampling theorem, the sample frequency in gatherer process is set to 50Hz.This experiment of the tested voluntary participation of all reservations, healthy, acardia disease and mental sickness medical history.Feelings induce material is evaluated as the cultural video fragment that can excite beholder's emotional responses.
2. data prediction and feature extraction
(2-1) training data intercepts
Sample database has preservation 175 sample datas altogether, and happiness, sadness, fear, indignation and tranquil five kinds of emotions that its chinesization is experienced often are planted and had 35 sample datas.Each sample data is generally at 5-10 minute, and wherein some time period excites expection emotion, and other times section mostly is tranquil emotion.Recording monitor is had to intercept 15 seconds long emotion fragments to each sample according to each tested whole process.These emotion fragments totally 175, glad, sad, frightened, angry and tranquil each 35 of five kinds of emotions.
(2-2) data smoothing filter is made an uproar
The spectrum distribution of GSR signal is within the scope of 0.08-0.2Hz, lower than most of interfering signal frequency, can not overlap with the frequency spectrum of other physiological signals and machine noise.Therefore and filtering smoothing to original GSR signal is only needed during research.Adopt long average window such as 25 hamming window functions smoothing, adopt Batterworth low pass filter filter out-band external noise, exponent number is set to 2, and cut-off frequency is set to 0.3Hz.As depicted in figs. 1 and 2, be the comparison diagram before and after one section of skin pricktest signal denoising respectively, the left side is before filtering, and the right is after filtering.
(2-3) physiological signal data normalization
Because galvanic skin response foundation level individual variation is very large, different application on human skin electricity levels is different, and same person also can be different under different time varying environment, so after intercepting data and filter makes an uproar, will be normalized data.The object be normalized data is to eliminate individual variation to a certain extent.The method of process is:
Wherein, the data of intercepted samples are expressed as X
gSR, the meansigma methods of these data is expressed as X
mean, maximum is X
max, minima is X
min, after normalization, data are expressed as X.
(2-4) feature extraction
18 features gone out as listed in table 1 to be extracted each intercepted samples, obtain 18 eigenvalues.
(2-5) characteristic value normalization
After obtaining 18 eigenvalues for each intercepted samples, there will be in 18 eigenvalues, some numerical value is very large, and some numerical value is very little, allly must be normalized eigenvalue.That the proportion that makes each feature shared in Emotion identification is equal to the object of characteristic value normalization process.Processing method is as follows:
Characteristic value normalization is normalized 175 eigenvalues corresponding to a certain feature of 175 intercepted samples.Its method is:
Wherein, primitive character Value Data is expressed as Y, and eigenvalue meansigma methods is expressed as Y
mean, maximum is Y
max, minima is Y
min, after normalization characteristic value data be expressed as Y ~.
3. use the emotion recognition improving k nearest neighbor method
Based on above several steps, we obtain 18 eigenvalues of 175 samples respectively.We will, based on these eigenvalues, utilize improvement k nearest neighbor method to carry out emotion recognition classification now.
For each test process, its concrete steps are, try to achieve the Euclidean distance of test sample book and 175 training samples; In 175 Euclidean distances obtained, choose 5 minimum Euclidean distances; Find out all kinds of emotions of training sample corresponding to these 5 Euclidean distances; According to the emotion of 5 training samples corresponding to these 5 Euclidean distances, and the size of these 5 Euclidean distances, obtain the classification that test sample book belongs to emotion.
In order to k nearest neighbor method is better described, Fig. 3 is the sample figure of k nearest neighbor method.In feature extraction for each sample extraction 18 features, so need the locus can showing each sample in the spaces of 18 dimensions, this is obviously unpractical.So this sample only employs first-order difference meansigma methods, second differnce evaluation of estimate, second differnce minima, three features show the emotion position in space that five kinds of culture are experienced.Wherein diamond symbols represents that the position at the sample place of tranquil emotion is experienced in culture.As can be seen from Figure 3 based in these three features, tranquil emotion has good discrimination, so be feasible by k nearest neighbor identification emotion.
According to method above, the emotion skin pricktest signal under experiencing 50 culture carries out emotion recognition experiment, and its result is as table 2.
Table 2: the recognition effect improving k nearest neighbor method
Result shows by experiment, and skin pricktest signal packet contains abundant emotion information, and the emotion recognition research adopting skin pricktest signal to be used for culture experience is feasible; Further, the effect that the k nearest neighbor method of improvement is used for emotion recognition is considerable, demonstrates the feasibility of the method.
Claims (2)
1. use skin pricktest signal to carry out a method for culture experience emotion recognition, it is characterized in that:
Described algorithm comprises the steps:
S1: obtain happiness, sadness, indignation, fear, this 5 class emotion physiological signal tranquil that culture is experienced by experiment;
S2: to the skin pricktest gathered and the smoothing noise suppression preprocessing of electrocardiosignal, remove the signal drift because collecting device and physical activity cause and burr;
S3: use basic signal processing method to extract 18 required in the method numerical characteristics of each sample; Relevant information is as shown in the table:
S4: the affective characteristics of each sample of labelling, is used for the k nearest neighbor model of cognition of emotional experience identification, for follow-up emotional semantic classification identification by the set composition being no less than 175 emotion physiological signals.
2. the method using skin pricktest signal to carry out culture experience emotion recognition as claimed in claim 1, is characterized in that: for the new emotional semantic classification identification gathering individual physiological signal, identification step is as follows:
S1: adopt skin pricktest signal pickup assembly to gather the skin pricktest signal of user when watching cultural product;
S2: the eigenvalue calculating the skin pricktest signal collected;
S3: the Euclidean distance calculating each sample character pair in all physiological features of test sample book and model of cognition;
S4: in all Euclidean distances, chooses 5 Euclidean distances that numerical value is minimum;
S5: the type of emotion of the training sample that these 5 Euclidean distances are corresponding in statistical identification model;
S6: calculate the ownership probability of tested sample for all kinds of emotion respectively;
Its computing formula is:
wherein θ
ifor current emotion is for the ownership probability of i-th kind of emotion; { 1,2,3,4,5} is corresponding glad, sad, indignation, frightened, tranquil five kinds of affective styles respectively for i ∈; K is the sample size belonging to the i-th type in 5 training samples nearest with test sample book; S is corresponding Euclidean distance numerical value;
S7: according to maximum ownership probability judgment emotion home type.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510528090.6A CN105212949A (en) | 2015-08-25 | 2015-08-25 | A kind of method using skin pricktest signal to carry out culture experience emotion recognition |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510528090.6A CN105212949A (en) | 2015-08-25 | 2015-08-25 | A kind of method using skin pricktest signal to carry out culture experience emotion recognition |
Publications (1)
Publication Number | Publication Date |
---|---|
CN105212949A true CN105212949A (en) | 2016-01-06 |
Family
ID=54982494
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510528090.6A Pending CN105212949A (en) | 2015-08-25 | 2015-08-25 | A kind of method using skin pricktest signal to carry out culture experience emotion recognition |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105212949A (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106691430A (en) * | 2016-12-21 | 2017-05-24 | 深圳欧德蒙科技有限公司 | Emotion judging method and device |
CN106691475A (en) * | 2016-12-30 | 2017-05-24 | 中国科学院深圳先进技术研究院 | Emotion recognition model generation method and device |
CN108670277A (en) * | 2018-06-04 | 2018-10-19 | 新华网股份有限公司 | Stress monitoring method |
CN108693974A (en) * | 2018-05-11 | 2018-10-23 | 新华网股份有限公司 | Data processing method, system and non-volatile computer storage medium |
CN109119162A (en) * | 2017-06-23 | 2019-01-01 | 永长欣医院管理顾问股份有限公司 | risk management system |
CN109199409A (en) * | 2017-07-06 | 2019-01-15 | 新华网股份有限公司 | Method and device for acquiring human body fatigue value |
CN110025323A (en) * | 2019-04-19 | 2019-07-19 | 西安科技大学 | A kind of infant's Emotion identification method |
CN113069115A (en) * | 2021-03-09 | 2021-07-06 | 清华大学 | Emotion recognition method, electronic equipment and storage medium |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20030139654A1 (en) * | 2002-01-23 | 2003-07-24 | Samsung Electronics Co., Ltd. | System and method for recognizing user's emotional state using short-time monitoring of physiological signals |
CN103315753A (en) * | 2013-06-24 | 2013-09-25 | 西南大学 | Portable sentiment analyzer and sentiment analysis method |
-
2015
- 2015-08-25 CN CN201510528090.6A patent/CN105212949A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20030139654A1 (en) * | 2002-01-23 | 2003-07-24 | Samsung Electronics Co., Ltd. | System and method for recognizing user's emotional state using short-time monitoring of physiological signals |
CN103315753A (en) * | 2013-06-24 | 2013-09-25 | 西南大学 | Portable sentiment analyzer and sentiment analysis method |
Non-Patent Citations (3)
Title |
---|
周钰婷: "皮肤电信号的情感特征提取及分类识别研究", 《万方数据学位论文》 * |
程静等: "基于皮肤电导的非线性情感特征提取研究", 《西南大学学报(自然科学版)》 * |
闫凤凤: "有序加权平均算子用于情感调节素材的选择", 《万方数据学位论文》 * |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106691430B (en) * | 2016-12-21 | 2019-10-11 | 深圳欧德蒙科技有限公司 | A kind of emotion judgment method and apparatus |
CN106691430A (en) * | 2016-12-21 | 2017-05-24 | 深圳欧德蒙科技有限公司 | Emotion judging method and device |
CN106691475A (en) * | 2016-12-30 | 2017-05-24 | 中国科学院深圳先进技术研究院 | Emotion recognition model generation method and device |
CN106691475B (en) * | 2016-12-30 | 2020-03-27 | 中国科学院深圳先进技术研究院 | Emotion recognition model generation method and device |
CN109119162A (en) * | 2017-06-23 | 2019-01-01 | 永长欣医院管理顾问股份有限公司 | risk management system |
CN109199409B (en) * | 2017-07-06 | 2021-06-29 | 新华网股份有限公司 | Method and device for acquiring human body fatigue value |
CN109199409A (en) * | 2017-07-06 | 2019-01-15 | 新华网股份有限公司 | Method and device for acquiring human body fatigue value |
CN108693974A (en) * | 2018-05-11 | 2018-10-23 | 新华网股份有限公司 | Data processing method, system and non-volatile computer storage medium |
CN108693974B (en) * | 2018-05-11 | 2021-09-21 | 新华网股份有限公司 | Data processing method, system and non-volatile computer storage medium |
CN108670277A (en) * | 2018-06-04 | 2018-10-19 | 新华网股份有限公司 | Stress monitoring method |
CN110025323A (en) * | 2019-04-19 | 2019-07-19 | 西安科技大学 | A kind of infant's Emotion identification method |
CN110025323B (en) * | 2019-04-19 | 2021-07-27 | 西安科技大学 | Infant emotion recognition method |
CN113069115A (en) * | 2021-03-09 | 2021-07-06 | 清华大学 | Emotion recognition method, electronic equipment and storage medium |
CN113069115B (en) * | 2021-03-09 | 2022-11-11 | 清华大学 | Emotion recognition method, electronic equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105212949A (en) | A kind of method using skin pricktest signal to carry out culture experience emotion recognition | |
Bota et al. | A review, current challenges, and future possibilities on emotion recognition using machine learning and physiological signals | |
Zenonos et al. | HealthyOffice: Mood recognition at work using smartphones and wearable sensors | |
Zhao et al. | EmotionSense: Emotion recognition based on wearable wristband | |
CN108446635B (en) | Collaborative filtering recommendation system and method for acquiring preference with assistance of electroencephalogram signals | |
US6546378B1 (en) | Signal interpretation engine | |
CN112120716A (en) | Wearable multi-mode emotional state monitoring device | |
KR102277820B1 (en) | The psychological counseling system and the method thereof using the feeling information and response information | |
Wu et al. | Modeling perceived stress via HRV and accelerometer sensor streams | |
CN107024987A (en) | A kind of real-time human brain Test of attention and training system based on EEG | |
CN103584872A (en) | Psychological stress assessment method based on multi-physiological-parameter integration | |
Jerritta et al. | Electrocardiogram-based emotion recognition system using empirical mode decomposition and discrete Fourier transform. | |
Yang et al. | Physiological-based emotion detection and recognition in a video game context | |
Khalili et al. | Emotion detection using brain and peripheral signals | |
CN103412646A (en) | Emotional music recommendation method based on brain-computer interaction | |
CN111523601A (en) | Latent emotion recognition method based on knowledge guidance and generation counterstudy | |
Pathoumvanh et al. | Robustness study of ECG biometric identification in heart rate variability conditions | |
CN111920420A (en) | Patient behavior multi-modal analysis and prediction system based on statistical learning | |
Li et al. | Multi-modal emotion recognition based on deep learning of EEG and audio signals | |
Petrantonakis et al. | EEG-based emotion recognition using hybrid filtering and higher order crossings | |
Kalansooriya et al. | Affective gaming in real-time emotion detection and Smart Computing music emotion recognition: Implementation approach with electroencephalogram | |
Singh et al. | Emotion recognition using electroencephalography (EEG): a review | |
WO2022100187A1 (en) | Mobile terminal-based method for identifying and monitoring emotions of user | |
CN117883082A (en) | Abnormal emotion recognition method, system, equipment and medium | |
Akinci et al. | Emotion modeling using fuzzy cognitive maps |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20160106 |
|
RJ01 | Rejection of invention patent application after publication |