Speech Features for Discriminating Stress
Julião, M. D.
; Silva, J. S.
; Moniz, G. S.
; Ferreira, J.
; Batista, M.
Speech Features for Discriminating Stress, Proc Conf. on Telecommunications - ConfTele, Aveiro, Portugal, Vol. 1, pp. 1 - 1, September, 2015.
Digital Object Identifier:
Download Full text PDF ( 217 KBs)
VOCE aims at doing speech classification as stressed or not-stressed in real-time, using exclusively acoustic-prosodic features. We look for the best discriminating feature subsets from a set of 6285 features – 6125 features extracted with openSMILE toolkit and 160 Teager Energy Operator (TEO) features. For that purpose, we use a mutual information filter and a branch and bound wrapper heuristic with an SVM classifier
to perform feature selection. We analyse all the chosen feature sets (280) on the performance they provide concerning accuracy, but also true and false positive rates. The results show that the best features types for our application case are Audio Spectral, MFCC, PCM and TEO. We reached results as high as 70.36% for generalisation accuracy.