Speech Analysis using Relative Spectral Filtering (RASTA) and Dynamic Time Warping (DTW) methods

This work consists of analysis of speech using RASTA and DTW methods.
The analysis is based on the speech recognition. Speech recognition converts identified
words or speech in spoken language into computer-readable format. The first speech
recognition has been developed in the year of 1950s. The variation of speech spoken
by individual becomes the main challenge for the speech recognition. Speech
recognition has application in many areas such as customer call centers and as a
medium in helping those with learning disabilities. This work presents an analysis of
speech for Malay single words. There are three stages in speech recognition which are
analysis, feature extraction and modeling. The Relative Spectral Filtering (RASTA) is
used as the method for feature extraction. RASTA is a method that subsidized the
undesirable and additive noise in speech recognition. Dynamic Time Warping (DTW)
method is used as the modelling technique.