Efficient estimation of perceptual features for speech recognition

A number of studies have shown that a pair of perceptual eeective formants can be deened to capture most of the phonetic information present in vowels. Various methods of computing the eeective formant values were proposed. However, many of them depend on the accuracy of conventional formant estimation. In this work, we study methods of automatically estimating perceptual eeective formants without estimating the actual formant values and compare the results with the perceptually measured eeective formant values. The preliminary results show that the method is eeective in estimating the perceptual eeective formants. Classiication experiments using perceptual eeective formants as explicit features do not demonstrate any advantages. However, using the perceptual eeective second formant value as input to our formant estimation algorithm can help to correct up to 44% of the formant tracking errors.