A Qualitative Analysis of Feature Extraction Based Action Recognition Techniques

Action recognition is an interesting problem and has many applications like surveillance, sign recognition, gesture and emotion recognitions. Many solutions to the problem have been suggested by various researchers. In this article a comparative study is performed between such most popular techniques. Various techniques are analyzed which make use of complex models like Discrete Wavelet Transform, Speeded Up Robust Features (SURF), Scale Invariant Feature Transform (SIFT) and image moments. All these models are comparatively analyzed in terms of speed and accuracy. We infer that the moments based algorithm feature extraction method is most balanced in terms of accuracy and efficiency and useful for providing an appropriate quality of results.