Face Recognition Based On An Improved PSO Algorthm

Face recognition is one of the biometric identification technologies; it involves theknowledge of physiology, psychology, graphics, and other disciplines. Similar with otherbiometrics, face recognition technology takes the biometric of human faces for identificationand authentication. With its advantage of convenience, not easy to be stolen, and security, theface biometrics has received an extensive attention. This paper begins with the introduction ofcommonly used face recognition technology, covering some typical identification methods,classification methods and feature extraction methods. The feature extraction technology offacial biometrics is as a primary task based on research context and research direction of facerecognition. The experimental results show that the binary PSO algorithm for facial featureselection and recognition can get a good resolution and better performance.Firstly, we choice the ORL face database as benchmark, it is an open face database thatscholars often use it to research the face recognition technology. After an initial preprocessingof facial images, the facial features are extracted using Discrete Cosine Transform, making thefeatures of the image to focus in the upper left corner of the processed images. Finally, thebinary PSO algorithm is used the facial feature selection and then we analysis the experimentresults.This paper describes the basic PSO algorithm and shows its suitable application scope aswell as its limitations according to the algorithm characteristics. Due to the characteristics ofgrayscale image, we select an improved PSO algorithm, i.e., the binary PSO algorithm, forfeature selection on the facial image. It can be seen from the experiment that the recognitionrate has been greatly improved using the binary PSO algorithm for image feature selection.Finally, the experimental results for testing comparison on the binary PSO and GeneticAlgorithm show more clearly that the binary PSO algorithm is a good feature selectionalgorithm and can be used to extract the object with most representative features.