Signal classification based on frequency analysis using multilayer neural network with limited data and computation

Signal classification performance using both multilayer neural network (MLNN) and conventional signal processing methods are theoretically compared under a limited observation period and computational load. Signals with N samples are classified based on the frequency components. A comparison is carried out based on the degree of freedom of the signal detection regions in an N-dimensional signal space. As a result, the MLNN has a higher degree of freedom, and can provide a more flexible performance for classifying the signals than the conventional methods. This analysis is further investigated through computer simulations. Multi-frequency signals and a real application in dial tone receiver, are considered. As a result, the MLNN can provide a much higher accuracy than the conventional signal processing methods