Gait training is essential in rehabilitation as it promotes mobility, prevents falls, and increases independence in daily activities. However, accurate gait assessment in the context of telemedicine remains a challenge. Wearable inertial devices offer value by providing a minimally invasive means to accurately capture movement. This study aims to improve telemedicine-based gait analysis by introducing a method utilizing a class II-B medical device equipped with a chest-worn inertial sensor. To enhance the quality of acceleration signals, we designed and 3D printed an 18-face polyhedron for stationary calibration. We validated this method by comparing its performance to a marker-based motion capture system in a cohort of young healthy adults. The results showed an 82% reduction in the root mean square error (RMSE) of acceleration signals. Furthermore, the study showed statistical and practical significant improvements in spatial gait characteristics. Specifically, there was a 14% reduction in root mean square error for gait velocity, a 59% reduction in RMSE for step length, and a 53% reduction in RMSE for stride length. These improvements in the accuracy of spatial gait feature detection can enhance telerehabilitation and ultimately lead to improved patient outcomes.