Pulse Coupled Neural Network Application In Human Face Pattern Analysis Research

Artificial neural network (ANN) is a hot subject in the field of information science in recent years, and it is a adaptive and nonlinear information processing system which is made of a number of the internet processing unit. In recent years, simulation the cognitive process of human brain becomes the focus of artificial neural network. People try to construct a step closer to the human intelligence information processing system to solve the problems existing in real life and science.Pulse coupled neural network (PCNN) is different from the traditional has a biology background and a new generation of artificial neural network. PCNN is closer to the real biological neural network in the activities of the nerve cells. It is a network with self supervision and self learning, so it does not need for advanced training. More importantly, its time series (a kind of statistical sequence of the number of pulse issued when in PCNN iterative calculation) has the effect of image feature extraction, and has the outstanding features of rotation invariance and scale invariance, etc. With the development of artificial intelligence, more and more researchers are paying attention to PCNN in recent years. It has been applied in image segmentation, image noise reduction, target recognition, image retrieval and other fields, and shows its unique advantages.In this master’s thesis, PCNN models were studied to image feature extraction, and were used in target face detection in color images and face expression recognition. Mainly to solve the problems of face recognition dependent on face the training set excessively and the difficult of expression feature classification, further extend the applications of PCNN in image feature extraction and face pattern analysis.Research work and new contributions of the dissertation are as follows:First, the dissertation summarized the PCNN model and principle, improved PCNN model and principle, simplified PCNN model and principle.Second, analyzed the principle and application of PCNN in image processing fields.Third, a method of target face detection using PCNN and skin color model was presented in this master’s thesis. We utilize a mixture skin color model to detect skin color areas, and screened the face candidate areas (sub-graphs) from the skin color areas. Then obtain the time series of benchmark face and all face sub-graphs using PCNN, and finally determine the target face in the tested images by computing the correlation of the time series between benchmark face and each sub-graph. Experimental results show that this method can rapidly detected the target face corresponding benchmark face independent of the face training set, and shows better detection performance.Fourth, the ignition location information of each iteration output of PCNN can fully reflect the image gray distribution and the details of image characteristics. According to this characteristic, a method of face expression recognition based on PCNN was presented in this master’s thesis. First use the average time series of PCNN to realize identity recognition. And then complete face expression classification by computing the cosine distance of ignition position sequences between the tested image and the seven expression images of the identity. Experiment and simulation show that ignition position sequence of PCNN can effectively extract face facial details features, and have good robustness.