A Multi-resolution Action Recognition Algorithm Using Wavelet Domain Features

This paper proposes a novel approach for human action recognition using multi-resolution feature extraction based on the two-dimensional discrete wavelet transform (2D-DWT). Action representations can be considered as image templates, which can be useful for understanding various actions or gestures as well as for recognition and analysis. An action recognition scheme is developed based on extracting features from the frames of a video sequence. The proposed feature selection algorithm offers the advantage of very low feature dimensionality and therefore lower computational burden. It is shown that the use of wavelet-domain features enhances the distinguish ability of different actions, resulting in a very high within-class compactness and between-class separability of the extracted features, while certain undesirable phenomena, such as camera movement and change in camera distance from the subject, are less severe in the frequency domain. Principal component analysis is performed to further reduce the dimensionality of the feature space. Extensive experimentations on a standard benchmark database confirm that the proposed approach offers not only computational savings but also a very recognition accuracy.