【摘要】 为解决在大数据下基于协作表示的快速人脸识别问题,提出一种通过选择各类近邻达到类内平均值最相似的局部协作表示算法。在原始训练集上,选择各类与测试样本近邻的若干样本,将各类近邻样本平均值进行稀疏重构;将重构误差最小的若干类近邻局部样本协作表示,实现分类。该算法通过较少样本的协作表示在提高识别率的同时,进一步降低了运算的复杂度。仿真实验验证了所提方法运用于人脸识别中的有效性。更多还原

【Abstract】 To solve the problem of fast face recognition based on the collaborative representation under big data,the collaborative representation algorithm which achieved the maximum similarity of the intra class mean by selecting the nearest neighbors in each class was proposed.A certain number of training samples which was nearest to the testing sample in each class was selected from the original training set,and the mean of the nearest neighbor samples in each class collaboratively represented for reconstruction.According to the minimal reconstruction errors,the nearest neighbor local samples which belonged to a certain number of classes was collaboratively represented to realize the classification.This algorithm improves the recognition rate,and further reduces the complexity of computation through the collaborative representation of fewer samples.The results of the simulation experiments indicate the proposed method is effective in face recognition.更多还原