Face Pose Estimation Based On Kernelized Maximum Separability

With the development of scientific technology and biometric technology,face recognition technology is widely used in all aspects of life and work.However,in the actual face recognition process,the recognition rate of face recognition is often reduced due to interference from external factors,such as changes in illumination,pose,expression.As an important factor that affects the performance of face recognition algorithms,the change of pose is also a direct factor that leads to a decline in the recognition rate of face recognition.Therefore,in the fields of computer vision,artificial intelligence,virtual reality,image processing,and human-computer interaction,facial pose estimation has become a hot topic.The Principal Component Analysis(PCA)method is an algorithm for solving face pose estimation,which can extract linear features in face images,construct a feature face space,realize data dimensionality reduction,and thus perform face pose estimation.However,PCA can only extract the linear features in the face image.The linear inseparable features in the face image cannot be extracted,and the face images distributed under the change of posture are highly nonlinear and complex.In order to solve this problem,a kernel function is introduced in this paper,and a face pose estimation algorithm based on kernel maximum separability is proposed.The experimental data based on CMU PIE and UMIST face database show that the algorithm has a high level performance of face pose estimation.In order to extract the nonlinear features in the face image and further estimate the face pose,the work of this paper mainly focuses on the following three aspects:(1)First,assign a face image in the face database to a face pose template.The purpose of this is to hope that the maximum variance of each pose template of the face image.(2)The kernel function as a good auxiliary tool to describe the nonlinear distribution can make us solve the problem of nonlinear feature extraction in face images.(3)After introducing the kernel function,a face pose estimation algorithm based on kernelized maximum separability is proposed.The experimental results on the CMU PIE and UMIST face databases show that the algorithm has a high accuracy rate in face pose estimation.The algorithm proposed in this paper involves the projection of kernel subspace,and the classifier used is the nearest neighbor classifier.