The Research Of Pulse Signal In Emotion Recognition

Affective Computing is an important branch of human-computer interaction research, which is on what is emotion, emotion how to generate and which aspects of emotion affected. Its aim is to give the computer to identify, understand and express the capacity of human emotions. The emotional state is a key research in affective computing, and the key to the physiological signal emotional state recognition is could distinguish between different emotional states features.The pulse signal is the research object. The paper extracts the main pulse signal wave position, and then extracts the pulse signal's statistical features in the happy, surprise, disgust, sadness, anger, fear of six emotional states, then utilizes calm feature for normalization. In this paper, genetic algorithm and Fisher classifier are used in dealing with the emotional state of the pulse signal features optimization problems.The paper is progressed the emotional state recognition for pulse signal, its main contents are as follows:1) Pulse signal data acquisition:experiments by MP150 collected through inspiring more than 300 subjects from with the state of happy, surprise, grief, disgust, angry, and fear of six states of the pulse physiological signals,it also collected the signal data in a calm state.The date of calm state is mainly used for normalization of six emotional features.2) Original pulse date preprocessing: mainly through wavelet de-noising function and using the filter to remove burrs of the original pulse signal and smooth the original pulse signal.The main case of the unsmooth pulse signal are:power signal interference mainly concentrate in the high frequency part, limit drift, and for human body shaking to generate such interference factors.3) Extraction of emotional features of pulse date:to discrete wavelet transform on the processed pulse signal, reconfigurate the high frequency part of signal pulse through adaptive threshold method to find the pulse signal wavelet peak position, and record by the position of the pulse-wave's dominant-wave,extract 104 the relevant statistical features.4) Selection of emotional features of pulse Subset: this essay by combining the genetic algorithm and Fisher classifier method to select the pulse signal feature, including genetic algorithm based on iterative times as estimate criteria function of evolutionary algorithm,to classify and recognite the six emotions, choose the effective combination which can stand for six emotional state.The emotional state recognition in one to one, the overall average of six emotional states training recognition rate can be reached 72.92%, the overall average verification recognition rate can also be achieved 69.68%, while the emotional state recognition in one to many, the six emotions state's overall average training recognition rate of can be reached 68.35%, the overall average verification recognition rate can be achieved 65.43%. This shows the emotional state of the pulse signal used for identification is possible, and can achieve certain effects.