The analysis of wearable motion sensors in human activity recognition based on mutual information criterion

Abstract

Selecting a suitable sensor configuration is an important aspect of recognizing human activities with wearable motion sensors. This problem encompasses selecting the number and type of the sensors, their position on the human body. In earlier works, researchers have used customized sensor configurations, and compared them with others in terms of the activity recognition rate. However, it is clear that these comparisons are dependent on the feature sets and classifiers employed. In this study, employing mutual information measure, sensor configurations are determined with respect to the time-domain distributions of the raw sensor measurements. The most informative axes of the accelerometers, gyroscopes, and magnetometers fixed at several locations on the human body are detected. © 2014 IEEE.