Speech feature extraction apparatus, speech feature extraction method and speech feature extraction program

摘要

A speech feature extraction apparatus, the apparatus comprising: a first difference calculation unit (600, 700, 800) for receiving a spectrum for each of a plurality of frequency groups of a speech signal, wherein the speech signal is segmented into frames for each frequency group, and calculating, for each frame of each frequency group, a difference of the spectrum between continuous ones Frame for the frequency group as a delta spectrum; and a first normalization unit (605, 710, 810) for performing normalization of the delta spectrum for each frame of each frequency group by dividing the delta spectrum by a function of the middle spectrum given by an average of spectrums across all speech frames.