การตรวจสอบการล้มในผู้สูงอายุโดยตรวจสอบรูปแบบการเปลี่ยนแปลงของจุดศูนย์กลางมวล

Abstract

This research presents the development of in-shoes multisensory for detected the abnormal gait in elderly people. As the cause that led to fall in the elderly. Numerous sensor types was installed in the shoes, such as the acceleration sensor, gyroscope, pressure sensors, bending sensor, which different features and performances in each type. Due to the complex of human walking, studies or measurements with a single sensor may not be enough. So, in this study took data from multi-sensory to study and analyze human gait, by divided into two phases, 1st when the foot touch the ground and 2nd when the foot do not touch the ground. Pressure sensors and bending sensor was applied to measurement, when the foot touches the ground. Data from pressure sensor are estimated Zero Moment Point (ZMP), ZMP are similarly every gait cycle of normal walking. The signal from bending sensor has small varying when human walk we cannot use to compare between normal and normal gait. During the foot not touching the ground, acceleration sensor was applied to study and classification, normal and abnormal gait. A three-axis accelerometer is installed at each foot to collect all three-axis data in Cartesian coordinate. Polynomial surface fitting technique is applied to the measured trajectory and classify. These data are subsequently transformed to spherical coordinate to form a 3-D gait trajectory. After studying the relationship of the signal from those sensors. In this research, artificial neural networks were used to study and classify normal and abnormal gait form pressure sensors signal, which is accurate about 90%. The 10 volunteers age between 18-25 year, high 150-175 cm. weight 40 - 75 kg that came for experiment. In the experimental of three-axis accelerometer and gyroscope data, around 74 gait cycles are tested and compare with simulation data, which is accurate about 85% of preliminary.