Frequency Extension of Telephone Narrowband Speech Signal using Neural Networks

In this paper we study the use of neural networks in bandwidth extension algorithms (BWE). The aim of such algorithms is to extend the frequency range of speech signals generating a large band signal from the corresponding narrowband signal, in order to improve its hearing quality. Based upon the presumption that the information contained in the available bandlimited speech and in the missing frequency components are correlated, we develop an algorithm which maps this relation using a neural network. The results were then analyzed and compared.