Development of a wearable free-weight exercise assistant

Abstract

In this work, the effectiveness of a system that uses sinusoidal motion models based on acceleration and orientation data to assess the quality of individual freeweight exercise repetitions was explored. Two inertial measurement units, one on each wrist, were worn by participants while performing correct and incorrect repetitions based on five common mistakes. Data were analyzed and relevant signals per exercise were selected. Based on readings from correct repetitions, the sinusoidal motion models were developed. The models were then coupled into three different systems that were evaluated based on the accuracy of counting repetitions and on the predicted quality of the repetition. The results depend on the system being evaluated, on the number and type of selected signals, and on the exercise carried out. Acceleration and orientation signals, when used together, yield an acceptable performance. For exercises without rotations, the sole use of acceleration data produces unsatisfactory results. Further work needs to be done before such a system can be used as a training tool with the purpose of improving exercising technique and help prevent injuries

Date of Award	30 Nov 2013
Original language	English
Supervisor	O.D. Amft (Supervisor 1), F.H. Wahl (Supervisor 2) & R.J. Bril (Supervisor 2)