Research On Intelligent Physiotherapy Apparatus Based On Expression Recognition

With the development of science and technology,electrical pulse physical therapy has received extensive attention and research in the field of physical therapy due to its advantages of no side effects and good effects.In recent years,with the rapid development of electronic medical equipment,the improvement of people’s quality of life and the increase of sub-healthy people,the existing physiotherapy products have been unable to meet people’s demand for better physiotherapy experience.In order to meet the needs of the quality of life,this article implements an intelligent physiotherapy system based on facial expression recognition for the current problems of low intelligence of physiotherapy products and poor physiotherapy experience.In this system,the upper computer transmits the real-time recognized expression information to the lower computer STM32 microcontroller,and realizes the function of modulating the pulse amplitude during electrotherapy by controlling the output signal.In order to realize an integrated physiotherapy platform with intelligent physiotherapy,this paper completes the design of the hardware system of the lower computer by designing hardware circuits such as system power supply circuit,voltage conversion circuit,and pulse modulation circuit.And by calling the facial expression recognition model to complete the host computer facial expression recognition system’s prediction of facial expressions.In order to improve the accuracy of facial expression information in controlling physiotherapy pulses.This paper optimizes the existing deep learning model mini-Xception to improve the accuracy of facial expression recognition.This optimization method uses the Mish activation function instead of the relu activation function,thereby improving the accuracy of model testing and accelerating the convergence of the model during the training phase.At the same time,the cross-stage hierarchical CSPNet structure is integrated into the mini-Xception backbone network,so that the architecture can achieve a richer gradient combination.This structure divides the feature map of the base layer into two parts,and then merges them through the CSPnet cross-stage hierarchical structure to fuse high-level and low-level features to improve the effect of target detection.Aiming at the stability of facial expression information to pulse control during physical therapy.This paper divides the 7 emoticons into 3 categories,and incorporates the median average filter algorithm into the control process of the lower computer to improve the stability of the system control.Median average filtering is a method of receiving the 3 types of expression digital sequences transmitted from the PC,processing,removing the maximum and minimum values in the data sequence,and then taking the arithmetic average of the remaining values to control the physiotherapy waveform Amplitude,so as to suppress random interference,and achieve the stability of electrical pulse signal control.