Other
One of the first decisions in any pattern recognition system is thechoice of what features to use: How exactly to represent the basicsignal that is to be classified, in order to make the classification algorithm's job easiest.
Speech recognition is a typical example. Through more than 30 yearsof recognizer research, many different feature representations of thespeech signal have been suggested and tried. The most popular feature representation currently used is theMel-frequency Cepstral Coefficients or MFCC.
Another popular speech feature representation is known as RASTA-PLP, an acronym for Relative Spectral Transform - PerceptualLinear Prediction. PLP was originally proposed by Hynek Hermansky asa way of warping spectra to minimize the differences between speakers while preserving the important speech information [Herm90]. RASTA is a separate technique that applies a band-pass filter to the energy ineach frequency subband in order to smooth over sh
matlab
No comment