Speech enhancement by noise cancellation using neural network

This paper describes a neural network speech enhancement system using a noise cancellation technique. Clean speech signals of each uttered digit in the Malay language are sampled from a single speaker in an almost noise free environment. Noisy speech signals are obtained by adding random noise to the clean signals. Noise cancellation is then performed on the noisy signals by using the ADALINE. Performance evaluation of the ADALINE is based on two methods: the signal-to-noise ratio (SNR) and by visual and audio checking. The effectiveness of the ADALINE to perform noise cancellation is also compared to the multilayer perceptron (MLP) network speech enhancement system.