A method of waveforms classification based on cascaded neural networks
Bendong Zhao, Shangfeng Chen, Huanzhang Lu, Junliang Liu · 2015
In this paper, a new cascaded neural networks approach of classifying waveforms is proposed. Compared to the traditional single back propagation networks (BP networks), the cascaded neural networks can decrease computational complexity and improve the classification performance at the same time in the case of the sheer volume of sampling data. The improved classification method adopts two-layered cascaded BP networks to realize the classification of waveforms. At first, the waveforms are partitioned into plenty of equal-length segments. Then every segment is coded using a numerical label according to its local structure features in the first layer networks. And in the other layer, the network is used to classify the combining sequences of segment codes. Experimental results show that the proposed scheme outperform other state-of-art BP classification methods in terms of the training efficiency and classification accuracy.