主成分分析与BP神经网络的人脸识别方法研究

Face Recognition Method of Principal Component Analysis and Back-propagation Neural Network

BP神经网络在人脸识别方面的研究中,原始样本数据不进行预处理与特征提取,不仅使识别结果准确降低,而且使BP神经的结构复杂化。主成分分析法能提取代替样本的少数几个主成分,这些主成分彼此不相关,符合特征优化的要求。BioID人脸数据库实验表明,将主成分分析与BP神经网络相结合,与传统单一的BP神经网络识别相比,提高识别的正确率,减少了训练时间,同时简化了网络结构,减少很大的计算量。

In the research of face recognition based on BP neural network the recognition precision will be low and the structure of BP neural network will become complex if sample＇s data is not preprocessed and features are not extracted. The principal component analysis can extract main factors that replace the whole face samples, furthermore these factors are not correlative each other and can well satisfy the features optimization. In BioID database experiment shows that firstly the principal component analysis used to process the face sample data,then the BP neural network used to recognize the face,compared with the tradition simple method,it improves the precision,reduces training time,simplifies structure of net and decreases the calculation.