SPEECH ANALYSIS WITH THE FAST CHIRP TRANSFORM (WedPmPO2)

Author(s) : Marian Kepesi (Vienna Telecommunications Research Center, Austria) Luis Weruaga (Austrian Academy of Sciences, Austria)

Abstract : The Chirp transform is a powerful analysis tool for variable frequency signals such as speech. The computational load represents the main limitation of its original formulation, discouraging its use in real-time applications. This paper analyzes a fast implementation, based on performing time-warping on the signal under analysis, combined with the Fast Fourier Transform. The performance of the Fast Chirp transform depends on the one hand on the estimation of the time-warping operation based on the signal characteristics, and, on the other hand on the interpolation technique used for the warping. Observations from the analysis of speech signals support the method and the further lines.

Menu