Singular spectrum analysis for gait patterns

Abstract

This paper proposes a new approach to gait pattern analysis based on acceleration signals during different walking conditions. Instead of applying traditional classification techniques, the proposed method looks into the characteristics of acceleration signals. Filtering and template matching methods based on singular spectrum analysis (SSA) and longest common subsequence algorithm (LCSS) have been used. The method has been used to discriminate walking downstairs, level walking and walking upstairs using 10 healthy subjects. The results suggest that the proposed method gives new insight into quantitative aspects of gait patterns.