仿生模式识别在单镜头人脸识别中的应用

A METHOD OF BIOMIMETIC PATTERN RECOGNITION FOR FACE RECOGNITION

提出了一种基于仿生模式识别(Biornimatic Pattern Recognition)和多权值神经元网络(Multi-Weights Neu-ral Network)的人脸识别新方法,对仿生模式识别理论在人脸识别中的应用模型作了讨论,并且介绍了一种新的人脸特征提取方法。本文通过实验对本文提出的基于仿生模式识别的方法和基于K近邻的方法做了对比,实验结果表明本文的方法克服了对未训练类型的人脸误识问题,提高了人脸识别系统的训练速度和正确识别率。

Biomimetic Pattern Recognition (BPR) is a new pattern recognition theory that was proposed by Wang Shoujue. The basic idea of BPR theory is based on the fact of the continuity in the feature space of any one of the certain kinds of samples. A new method of face recognition, based on BPR and Multi-Weights Neuron Network (MWNN), had been proposed. According to the theory of BPR, it is a continuous process when one's face turns from left to right (or right to left). Based on this idea, we can get a set of faces of a person from left to right in different angle in sequence by one camera, then a set of feature points in sequence of these faces can be got by face feature extraction. A space curve can be constructed by joining these points by line segments one by one in sequence. Considering some small changes on other directions, the covering shape of the person's face in the feature space can be regarded as a topological product of the one-dimensional curve and a 512-dimens,ional hy-persphere. That is the sub feature space of the face of the certain person in the feature space. A new method of facial feature extraction also had been introduced. The results of experiments with BPR and K-Nearest Neighbor Rules showed that the method based on BPR can eliminate the error recognition of the samples of the types that not be trained, the correct rate is also enhanced.