The Research And Verification On Key Technology Of Intelligent Terminal Human Gait Feature Recognition

In recent years,with the rapid development of smart terminals,the research direction of gait recognition technology is turning to computer vision to portable smart terminals.The smart terminal human gait recognition technology is not subject to scene restrictions,and is more in line with "anytime,anywhere" requirements,and is more suitable for promotion and application.Therefore,this thesis studied the key technologies of human gait recognition in smart terminals.The content of this thesis including:Firstly,this thesis studied the development and status of smart terminal human gait recognition technology.This thesis explained research significance and development prospect of gait recognition applied to intelligent terminal,investigated the status of related application products and key technologies,and summarized the existing problems at the current stage.Secondly,this thesis studied research focused on improving the recognition rate of gait movements in smart devices.This thesis firstly studied the gait acceleration cycles and types of humans,discussed solutions to reduce the impact of cell phone location moves,and then,summarized and analyzed the common gait features in Chinese and foreign literature,a total of 11 species,increased two new frequency domain features: spectral peak position,spectral center of mass,to constitute a new feature extraction scheme,finally,analyzed different evaluation methods and main indexes of the classification model,and determined the model evaluation methods and indicators of this thesis.Thirdly,this thesis designed a pattern recognition scheme and vertified by simulation focused on the key technologies of human gait recognition in smart terminals.This thesis firstly designed and developed a gait data acquisition software of Android mobile phone to get the original gait acceleration;then used a moving average filter to preprocess the data,extracted 13 kinds of feature values,and performed feature fusion and dimension reduction;then selected 4 classifiers used to identify and contrast experiments,and optimized the internal parameters of each classifier,finally,analyzed the recognition results of each classifier,and determined the optimal feature combination and classifier scheme based on the evaluation indexes.The simulation results show that the data volume is reduced by 53.8% by feature reduction,and the SVM recognition result is 94%,increased by 2% by the optimization of the classifier parameters.Fourthly,this thesis designed and developed a gait recognition software of Android smart phone based on the identification scheme of this thesis.The software can automatically identify the gait type,and has functions of step counting,distance statistics,heat statistics,track map,etc.Gait details records can be synchronized to the Bmob cloud data platform for viewing.Finally,in real-world testing,the average accuracy of software gait recognition is 93.2%,of which the running status is 100%,the walking status is 88%,the upstairs status is 84%,and the downstairs status is 94%.This thesis designes a scheme of gait feature extraction and recognition,vetifies the feasibility and optimization of the scheme by simulation,and developes a gait recognition application in smart phone base on the scheme.This thesis can provide reference for smart terminal gait recognition technology and application.