In this paper 'soft computing' algorithms for audio signal restoration are considered in regard to a practical digital sound library application. The methods presented are designed to reduce empty channel noise, being applicable to the restoration of noisy audio recordings. The audio signal is processed iteratively by a noise-reduction algorithm based on an intelligent comparator, improving the signal-to-noise ratio slightly at each iteration. At each time step, a fuzzy reasoning algorithm processes two values representing spectral power density estimates considered as linguistic variables. We describe a comparator module based on a neural network which approximates the distribution representing a non-linear function of spectral power density estimates. We demonstrate experimentally that the methods examined may produce meaningful noise reduction results without degrading the original sound fidelity. They have been applied to a practical Internet-based sound library (http://www. youarchive.net). © 2013 Springer-Verlag Berlin Heidelberg.
CITATION STYLE
Czyzewski, A. (2013). Intelligent control of spectral subtraction algorithm for noise removal from audio. Studies in Computational Intelligence, 467, 475–488. https://doi.org/10.1007/978-3-642-35647-6_28
Mendeley helps you to discover research relevant for your work.