000082782 001__ 82782
000082782 005__ 20180317093232.0
000082782 037__ $$aREP_WORK
000082782 245__ $$aEvaluation of Formant-Like Features for ASR
000082782 269__ $$a2002
000082782 260__ $$aMartigny, Switzerland$$bIDIAP$$c2002
000082782 336__ $$aReports
000082782 500__ $$aPublished: ICSLP 2002, Denver
000082782 520__ $$aThis paper investigates possibilities to automatically find a low-dimensional, formant-related physical representation of the speech signal, which is suitable for automatic speech recognition (ASR). This aim is motivated by the fact that formants have been shown to be discriminant features for ASR. Combinations of automatically extracted formant-like features and `conventional', noise-robust, state-of-the-art features (such as MFCCs including spectral subtraction and cepstral mean subtraction) have previously been shown to be more robust in adverse conditions than state-of-the-art features alone. However, it is not clear how these automatically extracted formant-like features behave in comparison with true formants. The purpose of this paper is to investigate two methods to automatically extract formant-like features, and to compare these features to hand-labeled formant tracks as well as to standard MFCCs in terms of their performance on a vowel classification task.
000082782 6531_ $$aspeech
000082782 6531_ $$aweber
000082782 6531_ $$abengio
000082782 6531_ $$abourlard
000082782 700__ $$aWeber, Katrin
000082782 700__ $$ade Wet, F.
000082782 700__ $$aCranen, B.
000082782 700__ $$aBoves, Louis
000082782 700__ $$0243961$$aBengio, Samy$$g140142
000082782 700__ $$0243348$$aBourlard, Hervé$$g117014
000082782 8564_ $$uhttp://publications.idiap.ch/downloads/reports/2002/rr02-04.pdf$$zURL
000082782 8564_ $$s101413$$uhttps://infoscience.epfl.ch/record/82782/files/rr02-04.pdf$$zn/a
000082782 909CO $$ooai:infoscience.tind.io:82782$$preport$$pSTI
000082782 909C0 $$0252189$$pLIDIAP$$xU10381
000082782 937__ $$aEPFL-REPORT-82782
000082782 970__ $$aweber-rr-02-04/LIDIAP
000082782 973__ $$aEPFL$$sPUBLISHED
000082782 980__ $$aREPORT