Signal classification using Bayesian regularization and Levenberg-Marquardt algorithm

This paper aims to investigate the neural networking system. The signals to be studied have been taken from photonic sensors. For classification, a given signal is first transformed into different feature domains and then neural network is used to train the given dataset to form the network. Wavelet transform is used to extract the signal properties-skewness, kurtosis and entropy and Fourier Transform is used to extract the energy of the signal. For training the data set which consists of basically 9 classes, two algorithms are used: Bayesian Regularization algorithm and Levenberg-Marquardt algorithm. The corresponding networks are made using training set and then these networks are used to classify the test set. Efficiency of the two algorithms is calculated to find out which is the more suitable one. They are then implemented to transfer an image in such a way that the trained network will decode the signal without the removal of noise.