Method Research On Emotion Recognition Of Surface EMG Based On Wavelet Packet And Wavelet Transform

With the rapid development of computer network communication and mufti-media technology, the technology of new Human Machine Interaction(HCI)has become a very active study subject in the computer science field at present.The study on the physiological signals emotion recognition has found important realistic values in such aspects as enhancing the intelligence and humanity of computer,developing new human-machine environments,promoting the study of psychology. Emotion recognition is a pivotal question of affective computing and a basic of founding harmonious human-computer interface environment.Efforts conducted by computational scientists have been mainly focused on three ways of detecting emotions:facial recognition,speech recognition,and a combination of the two(bimodal). Recently,greater attention has been paid to internal bodily manifestations,especially those related to the autonomic nervous system and the brain.Physiological signals represent the most promising and objective manner for detecting emotions in computer science.Furthermore,bodily signals can be recorded and analyzed based on algorithmic or auto-inference models that do not require human intervention.This paper adopts the physiological signals data to come from Augsburg University in Germany.This paper adopts physiological signals data is surface Electromyogram signals.This paper has mainly finished two research jobs:(1)Adopting the wavelet transform to analyse the surface EMG signal instability feature Surface EMG signal is decomposed by discrete wavelet transform(DWT)and selected maximum and minimum of the wavelet coefficients in every level.The extracted maximum and minimum of the wavelet coefficients is inputted to identify emotion by the BP neural network improved by Levenberg-Marquardt algorithm and classifier based on template matching of Minimum-distance. (2)Using the wavelet packet to analyse the surface EMG signal and extracting the wavelet packet coefficients to identify emotion.To confirm the wavelet packet ehtropy on classifying emotion through a large amount of simulation experiments.This paper did many simulation experiments,verified the feasibility and correctness of the above job,and obtained some corresponding results:(1)Overall classification results BP neural network classifiers better than nearest neighbor classifiers.a single physiological signal——joy,the recognition effect the nearest neighbor classifier better than the BP neural network classifier about on the emotional state of the surface EMG joy clustering effect.Verified the use of a single physiological signal can also achieve better recognition effect.If the BP network design and further improve the learning algorithm,as well as the wavelet transform feature extraction can be improved to improve the rate of emotional identification.The BP network is the future work of the network design and improved learning algorithm,as well as wavelet transform feature extraction improvements.(2)With the relative energy space wavelet packet characteristics as factor entropy is a good feature extraction methods can be more ideal recognition results.Wake up and explain the high degree of surface EMG relatively orderly,low wake of the Surface EMG comparison disorderly chaotic,random energy of the scattered in various frequency bands.Experimental results proved that the extract Surface EMG wavelet packet coefficient entropy analysis Surface EMG,and emotional state recognition is feasible and effective.But using wavelet packet coefficient entropy at the same time to identify the joy,anger,sadness and pleasure are not ideal emotional state effect should,be further study to find more effective features and join in or to study further from other physiological signals(such as ECG,respiration,EEG,etc)wavelet packet coefficient of the surface EMG entropy of wavelet packet coefficients entropy at the same time to identify with joy,anger,sadness and pleasure four emotional state.