Design Of Sports Motion Recognition System Based On Smart Insole

So far,in the field of sports,coaches have mainly relied on observing the performance of athletes to formulate suitable training plans for athletes.This has extremely high requirements for the professional quality of coaches.As people's enthusiasm for sports continues to rise in recent years,the number of professional coaches has become difficult to meet the demand of the majority of sports enthusiasts,and this phenomenon is more evident in small and medium-sized cities in China.Based on the above requirements,this paper designed a sports motion recognition system for sports enthusiasts based on smart insoles,and for the purpose of verifying the applicability of the system to different sports fields,experiments were conducted on basketball motions and race walking motions.The research contents and innovations of this paper include:(1)A basketball motion recognition scheme based on a convolutional neural network is proposed.First,one-dimensional median filtering,one-dimensional digital filtering and Z-score standardization are used to preprocess the sensor data,and then the performance of the neural network reaches the optimal value by adjusting its network structure and hyperparameters such as the size of the convolution kernel and the convolution step in the convolutional layer and the pooling layer.The experimental results show that the convolutional neural network can achieve an overall recognition accuracy of 99.3% in basketball motion recognition.(2)Since a basketball motion appears randomly in a whole period of actual basketball motion data and lasts for a random length of time,a method based on sliding window segmentation and adjacent multi-window joint judgment is proposed to recognize actual basketball motions.Experiments show that using this method to recognize actual basketball motions can achieve an overall recognition accuracy of 92.2%.(3)Convolutional neural network is used to detect race walking fouls.Experimental results show that this method is better than other motion recognition methods at capturing the slight differences between standard race walking motions and foul motions,thus obtaining an recognition accuracy of 98.1%.The above research contents show that the system can obtain a good recognition effect when recognizing basketball motions and race walking motions,which verifies that the system is an motion recognition system suitable for different sports fields.