Speech Enhancement in Hearing Aids Using Conjugate Symmetry of DFT and SNR-Perception Models

Most of the speech enhancement algorithms use the magnitude of STFT while phase is kept unchanged [3]. In this paper the magnitude of STFT of noisy speech is kept unchanged while the phase is modified. Modified complex spectrum of speech is obtained by combining unchanged magnitude spectrum and modified phase spectrum. This modification results into cancellation of low energy components (noise) of complex spectrum more than the high-energy (speech) components leading to reduction in background noise. SNR-Perception models for AWGN, Train noise, and Babble noise are developed using Nonlinear Regression modeling. Normal hearing and hearing loss subjective listening tests and spectrogram analysis show that the proposed method gives improved speech quality.