压缩感知理论打破了传统人脸识别模式,通过充分利用信号稀疏性采样理论,克服采样数据量巨大等问题.然而当人脸发生遮挡时,信号重构的准确性受到影响.针对这一问题,提出了改进的基于分块的稀疏表示方法,将遮挡图像进行分块处理.运用稀疏度自适应匹配追踪算法,来解决实际生活中信号稀疏度不定的情况,从而达到提高识别精度和鲁棒性等性能的目的.实验结果表明,改进后的算法能够很好的弥补发生遮挡时带来的不利影响,比传统的人脸识别方法的识别结果更加精确,运算速度更快.
The appearance of Compressed Sensing broke the traditional pattern of the Face Recognition. Using the sparse sampling theory, we have overcome problem of the great amount of sampling data. However, as a result of Face occlusion, the accuracy of signal reconstruction has been influenced. Aiming at this problem, an improved method based on block sparse representation is proposed, which is based on block processing. In order to address the situation in real life signal sparsity uncertain , using sparsity adaptive matching pursuit algorithm, achieve the purpose of improving the recognition accuracy and robustness properties. The experimental result show that the proposed algorithm can make up for the negative influence of Face occlusion, and more accurate compared with the traditional algorithm.