Research On Human Gait Recognition Based On The Fusion Of Plantar Pressure And Attitude Information

In recent years,human gait recognition systems have been widely used in gait detection,gait rehabilitation,walking assistance and other fields,but there are still key problems such as low comprehensive technical level,poor versatility and singularity of recognition strategy.Aiming at the above problems,human gait recognition based on the fusion of plantar pressure and attitude information is researched in this thesis.The main research contents are as follows:(1)Acquisition and analysis of gait information.Firstly,the foot pressure detection system based on MCP6004 amplifier circuit and STM32F103 signal processing module is designed.Secondly,BWT901 BLE attitude sensor and WT52 HB Bluetooth adapter are applied to set up the attitude angle detection system.Then the gait information is preprocessed by filtering and normalizing.Finally,the performance of the detection system is analyzed,which provides data basis for the research of gait recognition method.(2)Research on gait recognition method.A human gait recognition algorithm model based on Ts Fresh-RF feature extraction is proposed according to the attitude angle information obtained from the Inertial Measurement Unit(IMU).Experimental results demonstrated that the average classification accuracy of the proposed algorithm is 91.0%,which is significantly higher than the traditional gait recognition methods.For plantar pressure information,a threshold algorithm is used for gait phase recognition,and the experimental results show that the algorithm can effectively identify four gait sub-phases: pre-support,mid-support,post-support and swing phase.(3)Research on gait information fusion and prediction.In order to get the best prediction model by integrating the gait information,boosting algorithm is compiled to train the foot pressure and posture information respectively,and then a number of weak classifiers are fused by voting method.The experimental results reflect that Light GBM algorithm based on voting method achieves a higher recognition accuracy.At the same time,the Auto Regressive Moving Average Model is established.The prediction result can better match the change trend of the original gait data.Finally,a human-computer interface for gait information is developed to visualize the gait data,which provides a convenient platform for gait recognition optimization.To sum up,the information acquisition system of plantar pressure and attitude angle is firstly built in this thesis.Secondly,the gait recognition algorithms based on Ts Fresh-RF and threshold are proposed respectively.Then the foot pressure and attitude information is fused and predicted by voting method.Finally,the experimental study of human gait recognition and prediction algorithm is carried out to verify the effectiveness of the algorithm.The research results of this paper can provide important technical support for human-machine interaction information fusion and human gait recognition.