A Novel Feature Extraction Method on Activity Recognition Using Smartphone

In the last few years, research on human activity recognition using the built-in sensors of smartphones instead of the body-worn sensors has received much attention. Accelerometer is the most commonly used sensor of smartphone for the application. An important step in activity recognition is feature extraction from the raw acceleration data. In this work, a novel feature extraction method which considers both the distribution and the rate of change of the raw acceleration data is proposed. The raw time series liner acceleration data was collected by a smartphone application developed by ourselves. The proposed feature extraction method is compared with a previously proposed statistics-based feature extraction method using two evaluation methods: (a) distance matrix before clustering, (b) ARI and FM-index after clustering using MCODE. Both results show that the newly proposed feature extraction method is more effective for daily activity recognition than the previously proposed method.