摘要
针对传统Gabor特征维数高及训练样本少时识别率低的问题,提出一种基于分块Gabor特征融合的面部特征提取算法,引入隶属函数对人脸进行分类。提取人脸Gabor特征,对同一个方向多个尺度的Gabor特征进行融合,为充分利用面部器官对人脸识别的贡献,将Gabor特征脸按重要性进行分块,引入模糊隶属度概念对人脸进行分类。在标准人脸库上进行的实验结果表明,该方法有效提高了训练样本不足时人脸识别的正确率,改善了人脸识别系统的性能。
Considering that the high dimensional deficiencies of face feature extraction based on traditional Gabor wavelet transform and low recognition rate caused by lacks of training samples,a face recognition algorithm based on modular Gabor feature fusion was proposed,and the face images were classified by introducing membership function.The Gabor multi-directional and multi-scale features of face images were extracted,and the features in the same direction at different scales were fused.The Gabor feature faces were divided into chunks according to the importance to make full use of the contribution of facial organs to face recognition,and the face images were classified by introducing the membership function.Experiments were conducted in standard face database.Results show that the algorithm can effectively improve recognition rate when the training samples are insufficient and improve the performance of face recognition system.
出处
《计算机工程与设计》
北大核心
2017年第3期719-723,共5页
Computer Engineering and Design
基金
国家自然科学基金项目(61373055)
江苏省产学研联合创新基金项目(BY2013015-35)
