A MLP Network based Adaptive UWB Modulation Scheme Recognizer

This paper proposes an artificial neural network based ultra-wide band modulation scheme recognition algorithm. A three-layer MLP network is designed to obtain the recognition task. The statistical characterization parameters of the UWB signal are extracted as the input neuron parameters of the input layer, and the hidden layer of the MLP network has two layers. Experiments show that the MLP network with tansigmoidal neurons in the hidden layer and with linear neurons in the output layer can achieve good recognition performance. The probability of the correct recognition of the MLP network with proper transfer functions is higher than 95% at 10 dB SNR condition. Compared with the traditional pattern recognition and statistics judgment algorithms, the MLP recognizer can achieve higher correct recognition probability and can classify the modulation scheme automatically. The method may be widely used in the design of adaptive coding UWB systems