Research On Gait Recognition Based On Inertial Sensor

With the advent of the era of big data and artificial intelligence,the need for identification and verification is increasing,and the research on biometric identification has received widespread attention.Gait refers to how the human body walks.Due to the biological differences of individuals,different individuals have unique gait characteristics,so gait can be used for identification.Gait recognition,as a type of biometric recognition,is very different from iris,face,fingerprint,or other biometrics.This is mainly reflected in the fact that gait recognition can collect gait data in a non-intrusive way.Inertial sensors are inexpensive,small,and easy to carry.Therefore,the use of inertial sensors to collect gait information and then perform gait recognition is an important research direction in the field of gait recognition.Due to its powerful nonlinear modeling capabilities and feature extraction capabilities,deep neural networks have been widely used in many fields in recent years and have achieved very good performance improvements.The application of deep learning models in gait recognition has also made great progress.Deep learning models,especially network models based on Convolutional Neural Network(CNN),mainly use supervised learning to automatically extract features.Due to the use of a model structure different from the shallow network,the extracted features Generally has excellent classification performance.The main work of this thesis is to use the acceleration and angular velocity data to conduct research on gait recognition based on deep neural network.This thesis proposes two deep frame networks: the first network uses a convolutional neural network to extract gait features in gait acceleration and angular velocity signals,and then uses the attention mechanism to strengthen the gait features,and finally classifies them again;the other network first uses the Context Encoding Layer(CEL)to extract the correlation between the gait information and the Spatial Pyramid Pooling(SPP)layer to extract the multi-scale feature information.At the same time,our algorithm also utilizes the center loss function to increase the distance between classes,and finally implements classification.In the experiment,in order to better extract the gait features,the gait period is firstly extracted,and then slice the dataalong the time axis according to the gait period.After that,these sliced sample data was fed into the deep network for feature extraction.Experimental results show that the networks proposed in this thesis have certain advantages with regard to recognition performance compared to other algorithms or networks.