A Novel Foot-Forward Segmentation Algorithm for Improving IMU-Based Gait Analysis

The use of gait analysis techniques to identify health conditions has increased significantly in recent years. Among the many technologies available, inertial sensor-based solutions are one of the most popular due to their cost, accuracy, and portability. However, the accuracy obtained in gait analysis is primarily determined by a reliable segmentation of the steps. Most classical algorithms use some direct information extracted from accelerometers and gyroscopes, such as angles or signal magnitudes, and although they are effective for standard walking modes, they are very sensitive to unusual walking styles. In this sense, it would be desirable to obtain an effective and robust walking-style segmentation method. In this article, we analyze and compare a typical angular velocity-based algorithm for gait segmentation (AVGS), a commercial software for gait analysis (GaitUpLab), and a new algorithm proposed by the authors based on foot-forward displacement for gait segmentation (FoDiGS). The three algorithms have been evaluated under different walking-style tests using the Optitrack optical system (gold standard). The new proposed FoDiGS algorithm detects 96.9% of the 3205 steps analyzed and improves the gait parameter estimation, decreasing the mean relative error (MRE) by 20.89% against the AVGS algorithm and by 28.87% compared with the commercial system GaitUp. The results suggest that the proposed method, which does not require any previous training with a database or adaptive thresholds, provides an accurate segmentation method for different walking modes and outperforms other well-known methods.