Using Smartphone Sensors to Accurately Measure Ground Reaction Forces during Physical Activity - A Feasibility Study

Abstract

Evidence supports the use of external accelerometers to estimate ground reaction forces (GRFs). Modern smartphones come equipped with a tri-axial accelerometer and thus potentially provide a more cost-effective and convenient means to track GRFs. The ability of smartphone accelerometers to accurately estimate GRFs has not been established. PURPOSE: To compare the ability of smartphone accelerometers to other commercially available accelerometers in estimating GRFs during walking and running. METHODS: Nine subjects had a custom smartphone application downloaded onto their phones. This application allowed access to and recording of the smartphone’s raw accelerometer data. The subject’s smartphone (SP), as well as two other commercial accelerometers (A1 and A2) were placed on the most lateral aspect of the subject’s iliac crest. Subjects walked on an instrumented treadmill (measuring GRF) at a self-selected pace (SSW) for one minute and ran on the instrumented treadmill at a self-selected pace (SSR) for one minute. Using dynamic equations of motion, GRFs were estimated from accelerometer data collected from the three accelerometers (SP, A1, A2). The accuracy of the three accelerometers to predict GRF was calculated as the root mean square difference (RMSD) between resultant GRF from the instrumented treadmill and resultant GRF estimated from the accelerometers across the middle 30 s of the one minute SSW and SSR trials. RMSDs were normalized to average treadmill resultant GRF across the 30 second trials. Two separate one-way ANOVAs were used to assess differences in normalized RMSD between the accelerometers (SP, A1, A2) for both the SSW and SSR trials. RESULTS: There were no significant differences in normalized RMSD between the three accelerometer predictions of GRF during SSW (normalized RMSD - SP: 0.181±0.055, A1: 0.151±0.030, A2: 0.153±0.030, p=0.215) or SSR (normalized RMSD - SP: 0.280±0.087, A1: 0.228±0.036, A2: 0.233±0.043, p=0.126). CONCLUSION: Smartphone accelerometer predicted GRF data did not differ from that provided by previously validated commercial accelerometers. Smartphone accelerometers thus potentially offer a valid means of tracking GRFs through the design of appropriate applications.