Quantitative Gait Assessment With Feature-Rich Diversity Using Two IMU Sensors

The importance of gait analysis in medical applications, such as in rehabilitation, has been widely studied. Wearable sensors have gained popularity owing to their convenience of use in a flexible environment, while providing accuracy and reliability, in comparison with the gold standard system, i.e., motion capture. In this study, we proposed a framework for quantitative gait assessment using only two inertial measurement unit (IMU) sensors, while extracting maximum number of features. Decreasing the number of sensors negatively affects the performance of gait assessment. However, through comparison with a motion capture setup and previous studies, we verified the potential and limitations of our proposed framework toward providing a compact sensing system with feature-rich diversity for gait assessment. The results revealed that the temporal differences were 4.22±15.48 ms (mean ± S.D.) and −8.31±21.02 ms (mean ± S.D.) in the initial contact and toe-off events, respectively. Additionally, with respect to the spatial features, the stride length and heel vertical displacement were overestimated by an average of 7.72 cm and 2.22 cm, respectively. We successfully extracted 17 gait features from two IMUs located on the foot. We have also demonstrated that symmetry index feature can distinguish normal healthy subjects and subject with recent history of lower-limb injury, which is important for clinical research.