An Emotion Recognition Method Based On Improved Empirical Mode Decomposition And Characteristics Of Pulse Signal

Pulse signal of emotion recognition belongs to the physiologic emotion recognition, is a wide prospect application of computer pattern recognition technology, playing a more and more important role in the field of human-computer interaction. As a kind of physiological signals, the pulse signal contains abundant of physiological and pathological information, its pathological information has been widely used in medical fields very early, and under the development of computer technology the physical characteristics has been paid great attention by people.This paper designed a model used to implement real-time emotion recognition of pulse signal acquisition system, developed a real-time display of the pulse waveform, cross-platform PC software, and implemented the pulse signal acquisition, storage, emotional recognition, and other functions. It also provided a possible to achieve real-time man-machine interaction for the development of mobile devices. At the same time, through the Empirical Mode Decomposition preprocessing the collected pulse signal, combining with the characteristics of the pulse signal waveform, at the end extracted the important characteristics of the pulse signal, and built the relationship between happy, angry, calm, sad the four kinds. This paper mainly includes the following several aspects.First, according to the characteristics of pulse signal, used the reflection type photoelectric sensor to pick up the pulse signal, then with the help of hardware that amplifying circuit, amplified the reasonable pulse signal, and with the use of single-chip microcomputer to collect the signal, at the end the collected raw data were upload with USB, then through the PC software to display the original pulse data with figure, to complete the collected data,storage and display the function of pulse signal.Second, analyzed the nature of empirical mode decomposition, then discussed its advantages and disadvantages. To introduce the method into the research on emotion recognition of pulse signal, aim at the problems in the algorithm in the pulse signal, The paper raised two different kinds of improved algorithm based on support vector machine(SVM) endpoint continuation algorithm of genetic algorithm and adaptive set of Empirical mode decomposition( EEMD), and analyzed the advantages and disadvantages of the improved algorithms. Finally selected adaptive set of empirical mode decomposition algorithm as the basic core algorithm of the pulse signal.Third, through the adaptive set of empirical mode decomposition algorithm to analyze the characteristics of pulse waveform, this paper proposed a positioning pulse and heavybeat wave peak method. And aiming at the shortcomings of this method, some improvements are put forward. The experiment proved that the method on wave peak position exactly, and provided the basis of extract pulse signal feature extraction.Finally, the extraction of pulse signal under different emotional states the main wave peak and the statistical characteristic of heavy beat wave peak and the pulse signal of the main wave peak of approximate entropy, realized the recognition of different emotional states. And transplanted the application to the PC software, achieved the design of the emotion recognition system.