Terrain and Direction Classification of Locomotion Transitions Using Neuromuscular and Mechanical Input

Abstract

To perform seamless transitions in powered lower limb prostheses, accurate classification of transition type is required a priori. We propose a structure to detect direction (ascent or descent) and terrain (ramp or stairs) patterns when a person transitions from over ground to stairs or ramp locomotion. We compared electromyography (EMG) and accelerometry performance with an emphasis on sensor fusion for improving classification. Seven healthy subjects were recruited for this initial study. Data were collected with accelerometers and EMG electrodes on the dominant leg, while subjects transitioned from over ground to ramp (ascent and descent) and stair (ascent and descent) locomotion. Linear discriminant analysis and support vector machine approaches were used as classifiers using feature spaces of both sensor types. The results indicate that transitions are better classified as terrain type than direction type (p < 0.001), suggesting a terrain focused approach for an efficient structure. We also show that EMG and accelerometry data sources are complementary across the transitional gait cycle, suggesting sensor fusion for robust classification. These findings suggest that a terrain and direction focused classification approach will be useful for inclusion in classification approaches utilized in lower limb amputee samples.