Step Length and Step Width Estimation using Wearable Sensors

Abstract

According to multiple sources, a rising number of people in the United States have physical difficulties related to gait motion affected by injuries or neurological diseases. Current gait assessments used by physical therapists are subjective and restrictive, as they rely on the presence of a trained professional and equipment that are costly and not easily accessible. Newer studies have concentrated on the use of wearable sensors to solve these issues while still maintaining their accuracy and precision. Accurate methods to estimate gait parameters step length and step width using wearable sensors are needed since these parameters are used to determine if a person's walking pattern is symmetric and/or if the person has balance disorders. This study proposes methods to accurately estimate the gait parameters step length and step width using wearable sensors. Participants of this study walked on the Computer Assisted Rehabilitation Environment at two different speeds while using Inertial Measurement Units. A deep neural network is presented to estimate step length using features extracted from the wearable sensors signals. A step-length-based algorithm for step width estimation is introduced. Mean absolute errors of 0.2396 cm and 1.92cm were obtained f o r step length and step width estimation, respectively. Furthermore, results showed that at higher speeds the step length tends to be higher.