Activity recognition from accelerometer signals based on Wavelet-AR model
Zhenyu He · 2010
In this paper, a new Wavelet-AR feature for activity recognition from a tri-axial acceleration signals has been proposed. We use Wavelet Transform to decompose the raw accelerometer signals and obtain the decomposed signals that can efficiently discriminate the different activities. After that we build autoregressive (AR) model for the decomposed signals and extract the AR coefficients as features for activity recognition. As a consequence, Multi-class Support Vector Machines is used to distinguish different human activities. The average recognition results for four activities using the proposed Wavelet-AR features are 95.45%, which are better than traditional features. The results show that Wavelet-AR feature obvious discriminates different human activities and it can be extract as an effective feature for the recognition of accelerometer data.