独立成分分析应用于人脸识别中的几个问题

摘要

Independent component analysis(ICA) now is widely applied in pattern recognition domain, such as face recognition.Firstly, the image data can be preprocessed and dimension is reduced, and then the independent image basis can be got by ICA.Finally, face recognition is processed in the subspace. In the experiment, the influence of training samples, man labels and independent basis numbers to recognition ratio are compared and analyzed.%独立成分分析近年来广泛应用于人脸识别等模式领域.首先对人脸图像进行预处理降维,然后利用ICA算法获得人脸影像独立基成分,利用人脸影像独立基来构造一个子空间,最后利用待识别图像在这个空间上的投影进行人脸识别.针对训练样本个数,训练人数以及独立基数目影响识别率等三个问题进行实验,得出结果并进行分析.