Use of multiple change detection in pattern recognition using Relevant Vector Machine and moving sum average filter

In this paper, the performance of the RVM (Relevant Vector Machine) in classification of multiple non stationary power signal disturbances has been evaluated. Relevant Vector Machine (RVM), one of the newest approach to pattern recognition and machine learning. This paper proposes a new type of filter as moving sum average filter which is used for multiple change detection in power signals. In this work the moving sum average filter has been designed in a one cycle back fashion for localization of multiple fault and feature extraction. The extracted features are given as input for classification to RVM classifier. The result shows the effective classification of multiple non stationary power signals presented in this paper.