Audio Restoration Using Sound Source Modeling

This paper presents new propositions to audio restoration and enhancement based on Sound Source Modeling (SSM). The main motivation is to take advantage of prior information of generative models of sound sources when restoring or enhancing musical signals. We describe a case based on the commuted waveguide synthesis algorithm for plucked string tones and devise a scheme to extend the bandwidth of guitar tones. Then, we study the de-hissing of guitar tones and propose a scheme in which the bandwidth extension method is applied as a post-processing stage to a spectral-based de-hissing procedure. According to our experiments, this is effective for the reduction of common side-effects associated with spectral-based dehissing methods, such as musical noise and signal distortion.