Region-based amplification system of speech for listeners with hearing impairments

In this paper, we propose a region-based amplification system of speech for listeners with hearing impairments using the Discrete Wavelet Transform (DWT). The proposed system classifies the speech waveform into four main regions then amplifies each one of them in particular sub-bands according to their frequency components. The classification scheme segments the waveform into silence, voiced, unvoiced or mixed regions. The amplification process consists in applying a set of variable gain factors within the selected sub-bands that have been generated using an audiometric frequency-scale. The DWT is used in both classification and amplification steps. Experimental results show that the proposed system delivers high quality amplification of speech regions.