该方法利用人脸具有镜像对称的自然特性，依据奇偶分解原理，生成成镜像奇、偶对称样本，井利用人脸对称图像作为训练样本，再利用主分量分析(PCA)对训练样本进行二阶相关和降维处理，然后对处理后的样本进行ICA特征提取。理论和分析实验证明，该算法有效减线了人脸受到视角、光照、人脸表情、姿势变化等因素的最响，又增加了训练样本容量，减少了计算复杂度，同时有效解决了小样本问题，提高了识别率.(The method uses the natural characteristics of mirror symmetry of human face, generates mirror odd and even symmetrical samples according to the principle of parity decomposition, uses the symmetrical images of human face as training samples, then uses principal component analysis ( PCA ) to carry out second-order correlation and dimensionality reduction processing on the training samples, and then carries out ICA feature extraction on the processed samples. Theoretical and analytical experiments show that the algorithm can effectively reduce the face by the angle of view, lighting, facial expressions, posture changes and other factors, and increase the capacity of training samples, reduce the computational complexity, and effectively solve the problem of small samples, improve the recognition rate.)