Classification of waveforms using unsupervised feature learning and artificial neural network
Bendong Zhao, Shangfeng Chen, Junliang Liu, Huanzhang Lu · 2015
A novel method is proposed for the classification of waveforms, which takes full advantage of the local structures in time-domain waveforms. Specifically, the wave curves are divided into plenty of equal-length segments first. Then all of the segments are clustered and coded by using unsupervised feature learning methods. After that, the waveforms can be seen as a sequence of segment codes. Finally the waveforms are classified by means of a multi-layered perceptron (MLP) in which using the sequential codes as its input. Experimental results show that the waveforms are successfully classified by the proposed structure compared to the method that using MLP alone in terms of accuracy and efficiency.