基于ICA和BP神经网络相结合的掌纹识别

提出了一种基于独立成分分析(ICA,Independent Comment Analysis)和多层前馈(BP,Back Propagation)神经网络相结合的方法对掌纹进行识别.首先采用一种新的方法检测角点,得到掌纹图像的不变特征点,根据这些点校正掌纹方向并得到掌纹的感兴趣区域.对该区域采用定点快速ICA算法(FastICA),得到掌纹特征子空间,然后构建BP神经网络,并采用训练样本得到的掌纹特征进行训练,得到合适的权值.对香港理工大学掌纹数据库进行测试,与主成分分析(PCA,Principal Components Analysis)提取特征的方法进行比较,取得了较高的识别率. 

Palmprint recognition based on ICA and BP neural network

A novel method based on independent comment analysis(ICA) and back propagation neural network(BPNN) was proposed to solve palmprint recognition.First,a new method was used to detect corner points as the invariant feature points in palm images.Then the palm images were aligned by the points,and the region of interest(ROI) of the images was gotten.In the next place,palmprint features were extracted from the ROI by a fast fixed-point algorithm for ICA(FastICA).By means of this method,the original palmprint ROI images were transformed into a small set of feature space.Finally,the BPNN was trained by the feature spaces of the samples in the training set to get the appropriate weights.After the training,perform palm classification was applied.The method was tested on the Hong Kong Polytechnic University palmprint database.Experimental results show that method based on ICA achieves higher recognition rate than that based on principal components analysis(PCA).