Research On Gait Recognition Algorithm Based On Feature Extraction And Deep Learning

Gait recognition is an emerging biometric recognition technology that can identify people based on their characteristics while walking.Compared with biometric recognition technologies such as iris,fingerprints and human faces,gait recognition technology has some advantages such as unique,long-distance collection,non-contact and difficult to hide.It has become one of the research hotspots in the field of identity recognition.Traditional gait recognition methods are roughly classified into two categories: model-based methods and non-model-based methods.The model-based method is based on human gait sequences to model.The non-model-based method takes the entire silhouette of the human body as the research object for recognition.With the application of machine learning in various fields,deep learning methods are gradually applied to the field of gait recognition.In gait recognition,feature extraction and recognition classification are two key parts.In order to obtain a higher gait recognition accuracy rate,the innovations in this article can be summarized as follows:(1)An improved extreme learning machine algorithm is proposed.In this paper,based on the structural characteristics of the Local Coupled Extreme Learning Machine(LC-ELM),combined with the optimization strategy of the Particle Swarm Optimization(PSO)algorithm,an improved extreme learning machine algorithmLocal Coupled Extreme Learning Machine based on Particle Swarm Optimization(LC-PSO-ELM)is proposed.The algorithm uses the PSO algorithm to optimize the parameters,reducing the impact of random input parameters on the performance of the algorithm and improving the performance.(2)A gait recognition algorithm based on LC-PSO-ELM is proposed.The algorithm uses Gabor filter to extract gait features with different directions and different scales from the gait energy image firstly,and then uses Linear Discriminant Analysis(LDA)to solve the problem of excessive feature dimension.Finally,LC-PSO-ELM is used to classify the extracted gait features.Based on the CASIA-A and CASIA-B data sets,the recognition accuracy of the proposed method is compared with other current mainstream algorithms.Simulation results show that the proposed algorithm has good performance and performs well in cross-view gait recognition.(3)An ensemble gait recognition algorithm based on traditional gait recognition algorithm and Deep Learning algorithm is proposed.In the traditional gait recognition algorithm,feature extraction and recognition are generally separated.First,feature extraction from the gait images is performed,and then gait recognition.The calculation complexity is small and the training time is relatively short,but the recognition rate needs to be improved;Deep Learning algorithms can directly perform feature extraction and recognition on images.Due to the large scale of the image data to be processed and the large number of parameters that need to be trained,although a higher recognition rate can be obtained,the training time is longer.Therefore,this paper proposes an ensemble gait recognition algorithm,which integrates the traditional gait recognition algorithm and the deep learning algorithm PCANET,both algorithms can give full play to their advantages,which achieves a higher recognition accuracy.