Energy efficient activity recognition based on low resolution accelerometer in smart phones

Smart phone is becoming an ideal platform for continuous and transparent sensing with lots of built-in sensors. Activity recognition on smart phones is still a challenge due to the constraints of resources, such as battery lifetime, computational workload. Keeping in view the demand of low energy activity recognition for mobile devices, we propose an energy-efficient method to recognize user activities based on a single low resolution tri-axial accelerometer in smart phones. This paper presents a hierarchical recognition scheme with variable step size, which reduces the cost of time consuming frequency domain features for low energy consumption and adjusts the size of sliding window to improve the recognition accuracy. Experimental results demonstrate the effectiveness of the proposed algorithm with more than 85% recognition accuracy for 11 activities and 3.2 hours extended battery life for mobile phones.