Research Of Facial Expression Recognition Based On Gabor Wavelet And Particle Swarm Optimization Algorithm

Facial expressions can represent the state of a person, including his mood andcognition. It plays a very important role in the intelligent human-computer interactionsystem. Facial expression recognition technology research invo lves a number of areas,such as computer vision, image processing, affective computing, intelligentcontrolling and pattern recognition. It is a multi-disciplinary and intersecting topic andhas developed into a abroad studying hotspot in recent years. Facial expressionrecognition system mainly includes three aspects: the face image pretreatment, featureextraction and facial expression recognition. This paper focuses on the research of theface image feature extraction method, and facial expression recognition algorithm. Italso makes simulation experiments on the MATLAB R2011a platform to verify theeffectiveness of the proposed algorithm.Firstly, the paper will do the preprocessing on the input facial expression image.The first step, we determine the type of the image, if it is a color image, convert it tograyscale image. The second is the image histogram equalization, stretching grayvalues of all pixels uniformly to between0and255. After that the image will be cut,that is the scale normalization. The last is the extracting ROI (Region of Interest)region of the face image, reducing redundancy of the image, making it ready for thenext expression feature extraction.Secondly, according to the research on the facial expression recognition based onrelevant literature at home and abroad, we propose a new combining featureextraction portfolio strategy.It is based on Gabor wavelet transform and2DPCA(two-dimensional principal component analysis)+LBP (Local Binary Pattern) andparticle swarm optimization algorithm. After the preprocessing for a face imageperformed, the Gabor wavelet will transform it to extract the Gabor features; then use"2DPCA+LBP" approach to reduce the high dimensional Gabor features to the lowerdimension. Finally, the paper will use particle swarm optimization algorithm tooptimize the resulting expression feature. The strategy has the advantages of reducingdimensionality and relatively higher classification rate. It is applied to the face imagefeature extraction process in this paper to give expression feature vectors.Lastly, the paper makes simulation experiments based on the MATLAB R2011asimulation platform to illustrate the effectiveness of the proposed algorithm. Based on the JAFFE expression database, we will put our algorithm and commonly used Gaborwavelet+PCA, Gabor wavelet+2DPCA feature extraction algorithms into comparativetests based on the facial expression recognition system. The results show thatcompared to other methods the proposed algorithm does have better effect in terms offeature extraction.