Gait Recognition Algorithm Research Based On Frame Difference Energy Image

Along with the progress of the times, society and information, biometric technology has been widely used in intelligent monitoring and identification fields because of its security, stability, reliability, universality, carrying easily and other advantages. Compared with fingerprint, face, retinal and other biometrics, gait recognition as the most promising long-rang biometric, boasts the features of non-invasive, non-contact, hard to disguise and hide and etc. In recent years, gait recognition research has become a more popular topic in the field of intelligent video surveillance, computer vision, medical diagnostics, access control systems and etc. Gait recognition enjoys broad application prospects and practical significance, drawing great concern from researchers at home and abroad.Gait recognition is a biometric identification technology that recognizes people’s identity via people’s walking posture, which offers the potential for long-distance biometrics. The critical work of gait recognition is to extract gait features that can effectively express the continuity cycle movement and then uses the specific gait characteristics to design appropriate classifier to classify and identify them. Analyzing the frame difference energy image of gait, the paper found that the line quality vectors of the frame difference energy image can be extracted as the gait characteristics, and then reduce the dimension of feature and the amount of calculation, meeting the needs of real-time recognition. Finally, classify and recognize individual identity via the continuous Hidden Markov Models(CHMM). The main researches of the paper are as follows:Firstly, the stage of gait Pre-processing. Gait image preprocessing includes moving object detection, image segmentation, binarization, morphological processing, connectivity analysis, etc. After that, get a complete binary gait image sequences. In addition, we obtain the quasi-periodic information of gait image to prepare for the follow-up work.Secondly, the stage of gait Feature extraction. The study analyzed the frame difference energy image of gait and found that it can effectively express gait correlation information between adjacent frames and difference individuals have more obvious discrimination features. Therefore, they can be used as gait feature. Literature[50] extracts frame difference energy diagram as a gait feature directly. This method has better recognition performance, but higher computational complexity. On this basis, the paper design the gait recognition algorithm used the line quality vector of the frame difference gait energy image as gait feature. The method not only can reduce the dimensions of feature effectively and the amount of calculation but also can maintain a good characteristic distinction.Thirdly, the stage of gait recognition. Use CHMM for pattern classification. The processing includes CHMM modeling and identification. The simulation experiment has been conducted in CASIA Dataset A and CASIA Dataset B Gait Database, and the experimental results demonstrate the effectiveness of the algorithm.