Enhancement of coded speech signals using artificial neural network techniques

Artificial neural network techniques have the capabilities of modelling both linear and nonlinear functions. It is believed that some sort of linear as well as nonlinear relationships do exist between a set of noisy signals and the corresponding clean signals and this can be exploited by using the neural network models. Therefore, the aim of the research work presented in the paper was to investigate the use of the artificial neural network techniques for the improvement of the quality of coded speech signals. The noise of primary concern here is the vector quantisation noise which has been added to the original speech samples during the coding process