Face Recognition Simulation Integrating Local Block Reconstruction and Linear Regression
The relationship between facial features and categories is often nonlinear,especially in the case of large changes in lighting,posture,etc.,which makes it difficult for linear models to capture nonlinear facial micro fea-tures,resulting in poor classification results.Therefore,a linear regression method for local block reconstruction is pro-posed.Samples after feature extraction are processed in blocks.The reconstruction coefficient is calculated for the block training sample set with more feature energy.By increasing the proportion of features with more feature energy in the linear reconstruction,the proportion of features with less feature energy is reduced to achieve the noise reduction effect.Then the reconstruction coefficient is applied to the overall training samples to obtain more similar reconstruction samples,and the maximum similarity method is used to determine the category of the test samples.Ex-periments on AR,CMU PIE,and other face databases show that the proposed method has better classification results.
Linear regressionChunking reconstructionClassifierFace recognition
万方数据
维普
