Monitoring of human body running training with wireless sensor based wearable devices

Abstract

With the rapid pace of today’s society, work pressure, less exercise time, people began to pay more attention to their health. Walking and running have become the first choice of moderate exercise for many young people. The recognition of human running state based on wireless acceleration sensor will play an increasingly important role in the fields of motion detection, energy consumption evaluation and health care. It is of great significance to develop and design a kind of wearable multi-functional wireless sensor which can monitor the running state of human body. In this paper, a wearable human body monitoring system based on wireless acceleration sensor technology is proposed for real-time monitoring of daily running volume of human body. Hardware and upper computer design: stm32f405 is used as the main control chip, and ma8451q is used to collect human motion data. In this paper, aiming at the problem that three kinds of motion states of human body are easy to be confused and difficult to distinguish, based on the in-depth study of the complex structure mode and self similarity characteristics of non-stationary acceleration signal, a method of human body motion state recognition based on single fractal and multi fractal is proposed. In this method, the fractal dimension and the generalized dimension are used as the feature variables, and the correlation judgment method is used to distinguish and recognize different motion states. Experiments show the validity and feasibility of single fractal and multifractal in walking and going up and down three kinds of motion state recognition. On the basis of multifractal motion state recognition, this paper combines fractal theory with wavelet multiresolution analysis, and proposes a matrix fractal human motion state recognition method based on wavelet transform. The fractal matrix based on wavelet transform quantifies the fractal characteristics of the component signals of walking and going up and down in different wavelet scales, and then describes the complexity and self similarity of the original acceleration signals. Experimental results show that the average recognition rate of walking, jogging and fast running can reach over 93% under the premise of less prior information.