Study On Facial Expressions Recognition Based On Extreme Learning Machine

Facial expression recognition is one of the core content of human-computer interaction research, intended to understand people’s hearts activities and inner emotions through analyzing facial expression by computers, and achieving human-computer interaction more natural and more intelligent. After decades of development, the technology of facial expression recognition has been widely used in various fields,and the research of expression recognition has important practical significance.ELM algorithm in the process of determining the parameters of the network, without any iteration step, largely reducing the time it takes to adjust the network parameters. compared with others traditional training algorithm of SLFN, ELM algorithm is faster, better at generalization performance. Therefore, ELM used in the expression recognition, to improve recognition speed and accuracy.This article using facial expressions as the research object, mainly studies the common expression feature used and face recognition method to categorize and summarize, analyzing and comparing the advantages and disadvantages of various methods. Based on this, we put forward a facial recognition method based on extreme learning machine. The main innovations and contributions of this article are as follows:First, Expression regional positioning method based on binary and grayscale integral projection combine. Expression area is an area of eyes, eyebrows and mouth around it. By detecting and locating of expression area from the face has very important significance to facial expression recognition.Second, Fusion Gabor wavelet transform and PCA methods emoticons special clinic extraction. Using multi-scale, multi-directional two-dimensional Gabor transform expression feature extraction. Then use the PCA method to extract Gabor feature dimension reduction.And, the design of the classifier based on extreme learning machine. By using the limit characteristics of machine faster training speed of learning, design expression classifier based on extreme learning machine.Last, to test the performance of the proposed face recognition method, through a lot of facial expression image test and calculate the final recognition rate, and then with the other uses the same Gabor wavelet transform method expression feature extraction, but the classification of different facial expression recognition by method based performance comparison recognition rate. Finally, we will work through a lot of the standard facial expression pattern to test and test our system. Simulation results showing that the ultimate learning machine can effectively improve facial expression facial expression recognition rate.