Researh On Motion Signalacquisition And Recognition On MEMS Acceleration Sensor

The rapid automatic identification method and theory of human motion play the importantly theoretical and practical significance in many areas,such as medical care,behavior assistance,man-machine interface,domestic service and home robot.Compared with motion recognition based on machine vision,sensor-based motion recognition is more advantageous on the application of behavior assistance and monitoring,because it is faster and easier to be connected with the control system.With the rapid development of MEMS sensing technology,the research on the areas introduced above has been paid increasing attention.The main contents of this paper are as follows:Firstly,in terms of building a experimental platform,the SOC chip nRF51822 based on the SoftDevice protocol stack which can achieve ultra-low power Bluetooth BLE 4.0 is selected as the main control chip.Combined with the industry's first three-axis digital acceleration sensor LIS3 DSH embedding two independent programmable state machine,the action recognition of the slave computer hardware system platform,realizing the software designing on data acquisition,Bluetooth communication,the state machine programming,is set up which has been testified under environment of nRF51822.The PC and smartphone(above android 4.0)are used as the means of realizing the communication between the host computer and the slave one.In this paper,the main performance of the hardware system of the slave computer is tested,including the power of the Bluetooth module,the power consumption and the stability of the communication data.Moreover,basing on the condition above,eight kinds of acceleration signals of the actions of turning over and walking are collected.By studying the relevant data processing and feature extraction methods,the features of the acceleration threshold and time threshold are extracted to construct a state machine model to identify specific actions.Finally,the results of theory and simulation are used to program the embedded state machine LIS3 DSH,and the effect of action recognition is tested and summarized by experiments.The results show that the motion recognition based on MEMS three-axisdigital acceleration sensor LIS3 DSH has the advantages that gaining fast speed and excellent real-time properties.However,the accuracy of recognition depends on the structure of the state machine.