Long-term gait pattern assessment using a tri-axial accelerometer

In this article, we present a pervasive solution for gait pattern classification that uses accelerometer data retrieved from a waist-mounted inertial sensor. The proposed algorithm has been conceived to operate continuously for long-term applications. With respect to traditional approaches that use a large number of features and sophisticated classifiers, our solution is able to assess four different gait patterns (standing, level walking, stair ascending and descending) by using three features and a decision tree. We assess the algorithm detection performances using data that we retrieved from a validation group composed by nine young and healthy volunteers, for a total number of 36 tests and 12.5 h of recorded acceleration data. Experimental results show that in continuous applications the proposed algorithm is able to effectively discriminate between standing (100%), level walking (∼99%), stair ascending (∼84%), and descending (∼85%), with an average classification accuracy for the four patterns that exceeds 92% in continuous, long-lasting applications.