SVD-based portable device for real-time hoarse voice denoising

For pathological voices, hoarseness is mainly due to airflow turbulence in the vocal tract and is often referred to as noise. This paper focuses on enhancement of speech signals, assumed to be degraded by additive white noise. Speech enhancement is performed in the time-domain, by means of a low-order singular value decomposition that allows removal of the noise contribution while preserving the harmonic structure of the original speech signal. The approach was tested on synthetic data and applied to real data, concerning hoarse voices of cordectomised patients. Its simple structure allows a real-time implementation, based on C and assembler code, properly optimised on a DSP board. Hence, the method is suitable for portable device realisation, as an aid to disphonic speakers. It could be of help for diminishing the effort in speaking, which is closely related to social problems due to the awkwardness of voice.