가속도 센서와 자이로 센서의 융합을 통한 신경망 기반 보행 유형 판별에 관한 연구

Abstract

Recently, many countries including Korea are becoming an aging society and thus interests in health monitoring have been increasing. People are seeking more efficient and convenient methods for confirming their health status. In addition, IoT (Internet of Things) technology has been actively developed to make an instant diagnosis. This thesis suggests a gait pattern discrimination method by collecting gait data using an acceleration sensor and gyro sensor via wireless Internet such as WiFi. A build-in module with the acceleration and gyro sensors was inserted into a sole of both shoes and the straight distance was measured with usual walking. After collecting gait data including the x, y and z coordinates associated with the acceleration and gyro sensors via WiFi, the gait data are converted into sensor values based on the earth gravity and the acceleration data are converted into position values. These values are used for extracting one step for walking that is a graphical form. The suggested research demonstrates the development of the gait pattern discrimination method by using a simple neural network for judging gait (i.e. toe-in, normal, and toe-out). After measuring the sensor data of both shoes of several test subjects, a series of gait data for one step of walking is used for learning of a neural network model. Experimental results show that the suggested method for gait analysis can judge walking patterns as the toe-in, normal and toe-out gait. It is believed that this method can help people to get an early diagnosis if they have problems with their gait patterns.