Noise reduction which is also known as speech enhancement algorithm improves one or more perceptual aspects of noisy speech, most notably, quality and intelligibility. Speech quality is a measure of how clear, natural and free of distortion the speech is. This project deals with two noise reduction algorithms: 1) Binary Masking Algorithm (BMA) 2) Deep Neural Networks (DNN). Binary Masking Algorithm (BMA):This algorithm is used to identify speech dominated and noise dominated units, using which a binary mask is calculated and applied to the noisy input spectrum to get the noise suppressed spectrum.
Keywords : Ideal Binary Masking, Deep Neural Network, Noise Quality, Oracle, Intelligibility.