Perceptual time-frequency subtraction algorithm for noise reduction in hearing aids.

Abstract

Sensorineural hearing disorders are a major and universal community health problem. In many cases, hearing aids offer the only solution for people suffering from such disorders. Unfortunately existing aids do not provide any improvement in intelligibility of the signal when background noise is present. A hearing aid system should ideally simulate auditory processes including those aspects of the speech signal that are perceptually important. This work presents a new integrated approach to the design of a digital hearing aid, based on a wavelet transform, as well as a formulation of the temporal and spectral psychoacoustic model of masking. Within the model, the Perceptual Time-Frequency Subtraction (PTFS) algorithm is developed to simulate the masking phenomena and reduce noise in single-input systems. Results show that the use of the PTFS yields a significant improvement in speech quality especially in unvoiced portions. Additionally, the noise component during periods of silence has been attenuated by up to 20 dB. This new noise reduction method is expected to be applicable in a variety of applications, including digital hearing aids and portable communication systems (e.g., cellular telephones).