摘要

The duration of the vowel and the nasal was analyzed in the casual pronunciation of Hungarian words containing the sequence Vn.C, where '.' is a syllable boundary and C is a stop, affricate, fricative, or approximant. It was found that due to anticipatory coarticulation the duration of n is significantly shorter before fricatives and approximants than before stops and affricates.
A teaching algorithm was used to distinguish between stops/affricates and fricatives/approximants in VnC sequences. We used an approach to the classification of C by means of the support vector machine (SVM) and the properties of Radial basis RBF) kernel (using MATLAB, version 7.0). The results show close to 95% correct responses for the stop/affricate vs. fricative/approximant distinction of C, as opposed to about 60% correct responses for the classification of the voicing feature of C.

  • 出版日期2010-6

全文