Improving the perceptual quality of audio signal in terms of degree of listener fatigue, intelligibility is the primary need of any audio based smart devices. Audio enhancement is needed for the applications like voice biometrics, audio-to-text conversion etc. Number of stationary and non-stationary background noises adds with the audio signal while recording or transmitting them. Addition of background noise signal deteriorates the quality of audio which decreases the efficiency of audio based applications. Various variants of spectral subtraction methods have been used for enhancing the quality of audio signal by reducing the background noises. This paper enhances the quality of audio by estimating the noise followed by applying modified spectral subtraction method. Here, noise is updated by identifying the absence of human speech signal in a frame.

Plant diseases, AI, computer vision, disease detection, neural networks, support vector machines, k-nearest neighbors, agriculture.