Improvement in hoarse voice denoising for real-time DSP implementation

Voice hoarseness is mainly related to airflow turbulence in the vocal tract. It can be due to vocal fold paralysis, polyps, cordectomisation or other dysfunction, which alter regular speech production, and is commonly treated as a noise component in the speech signal. A denoising approach is proposed, based on low-order singular value decomposition (SVD) of matrices whose entries come from sampled speech data frames, properly organised. A prototype DSP board implementing the procedure was developed, by means of properly optimised C and Assembler code. Enhanced results are obtained with respect to a previous scheme, by introducing a normalization step on the signal amplitude dynamics. This allows increasing the output level, as well as reducing click-noise, both due to the algorithm structure and DSP implementation constraints. Objective quality indexes are proposed, showing the better results achieved with the proposed modifications.