Speech enhancement using auditory-based transform

Abstract

Auditory-Based Transform (AT) is a function based on the impulse response of the basilar membrane in the cochlea. As the human auditory system utilizes the time and frequency localization of audio signals, AT can be the best suited algorithm for speech enhancement that is for reduction of noise in speech signal processing. The present paper deals with the implementation and comparison of Auditory Transform with two other algorithms: -Spectral subtraction and Wiener filter. Evaluation of intelligibility and comparison among the three algorithms has been done based on a recently proposed objective parameter, Normalized Covariance Measure, along with PSNR and Spectrogram.