Performance comparison of wavelet families for analog modulation classification using expert discrete wavelet neural network system

This study presents a comparative study of implementation of feature extraction and classification algorithms based on wavelet decompositions and MLP for analog modulated communication signals. In this paper, F-db2, F-db3, F-db5, F-db8, F-db10, F-sym2, F-sym3, F-sym5, F-sym7, F-sym8, F-bior1.3, F-bior2.2, F-bior2.8, F-bior3.5, F-bior6.8, F-coif1, F-coif2, F-coif3, F-coif4, and F-coif5 feature extraction methods are generated by separately using db2, db3, db5, db8, db10, sym2, sym3, sym5, sym7, sym8, bior1.3, bior2.2, bior2.8, bior3.5, bior6.8, coif1, coif2, coif3, coif4, and coif5 wavelet filters. Then the performance comparison of these feature extraction methods is performed by using a discrete wavelet neural network (DWNN) expert system. The analog modulated signals used in this study are six types (AM, DSB, USB, LSB, FM, and PM). DWNN model is used, which consists of two layers: discrete wavelet-adaptive wavelet entropy and multi-layer perceptron (MLP) neural networks for expert analog modulation classification. The performance of this comparison system is evaluated by using total 1920 analog modulated signals for each of these feature extraction methods. The performance comparison of these features extraction methods and the advantages and disadvantages of the methods are examined. The rate of mean correct classification is about 95.81% for the sample analog modulated signals.