Wearable Human Activity Recognition by Electrocardiograph and Accelerometer

ABSTRACT:

               Recently, life log services are receiving considerable attention in medical and health management. The life logging for health management records biological information such as blood pressure, body weight and activity. This paper proposes a human activity estimation system using a wearable multi-sensor with a built-in electrocardiograph and triaxial accelerometers. The multi-sensor unconstraintly measures biological information, and provides these data to personal computer by wireless communication. We estimate human activity in a series of activities by the biological information. In our experiment, the subjects have several activities such as “Walking”, “Rest” and “Strength training”. The system estimates these activities by a decision tree. Branch conditions of the decision tree are aided by fuzzy logic and state of activity transition from previous activity. Fuzzy membership functions are constructed from exercise intensity, distinction frequency and transitional probability. As the results, the proposed method estimated activities with good accuracy.