[2I6-GS-2] Machine learning: Sensing and user aid

[2I6-GS-2-01] Behavior Analysis for Walking Motions Using Characterization of Accelerometer Data

〇Kosuke Shima¹, Atsuko Mutoh¹, Koichi Moriyama¹, Nobuhiro Inuzuka¹ (1. Nagoya Institute of Technology)

Keywords:Behavior Analysis, Activity Analysis, Acceleration Sensor, Wearable Computing

Humans usually perform various activities, in while, walking is one of important activity. The activities are performed with various motions. For example, humans walk with swinging their shoulder, swinging their arms, or watching smartphones. We assume these difference of motions are according to cultural, physical, and mental characteristics and conditions.

In this work, we proposed a method that observes human motion using accelerometer values. This method characterizes acceleration values to sequences of symbols robustly against timing and duration of motions. The sequences express sequence of motions. In this paper, we propose a method that observes human motion with the sequences using duration of motions as additional information. The proposed method observes not only differences of motions but also observes temporal differences at same motions.

In an experiment, we confirmed that the proposed method classified accelerometer values into amounts of steps in walking, temperature, and home prefecture.