Speech Enhancement Using Neural Network

This paper describes a neural network speech enhancement system using Multilayer Perceptron (MLP) network and trained using the back propagation algorithm (BPA). Speech enhancement is generally refers to map noisy speech into cleaner speech. Noisy speech signals are obtained by adding random noise to the clean signals .Speech enhancement is then performed on the noisy signals by using the ADALINE. Here we show that neural nets can be used to significantly boost recognition accuracy, without retraining the speech recognizer.