R. Vetter, P. Renevey, and J. Krauss (Switzerland)
Genetic algorithm, speech enhancement, perceptual constraints, multidimensional signal processing, coherence function
We present a method for the optimization of parametric speech enhancement methods by means of genetic algorithms (GA). It is illustrated on a dual channel speech enhancement method which operates in the perceptual do main and exploits the spatial diversity of the sound field by the coherence function. This algorithm necessitates the adjustment of some crucial parameters with regard to a specific end-user and/or application. Traditionally, this optimization problem is carried out by hand, involving a cumbersome and time consuming parameter tuning. In order to circumvent this laborious task we propose herein an optimization based on genetic algorithms.
Important Links:
Go Back
