An efficient method for activity recognition of the elderly using tilt signals of tri-axial acceleration sensor

We propose an activity recognition system for the elderly using a wearable sensor module embedding a tri-axial accelerometer, considering maximization of battery life. The sensor module embedding both a tri-axial acceleration sensor and an RF transmission module is worn at the right side of one's waistband. It connects and transfers sensing data to subject's PDA phone. Then, an algorithm on the PDA phone accumulates the data and classifies them as an activity. We utilize 3 tilts in addition to 3 acceleration values, compared to previous works. However, we reduce the sampling rate of the sensing data for saving battery life. As an activity classifier, the SVM (Support Vector Machine) algorithm is used, and we have achieved 96% of accuracy in detecting an activity out of 9. It shows the proposed method can save the battery life without losing the recognition accuracy compared to related works.