1st Joint SIG-IL/Microsoft Workshop on Speech and Language Technologies for Iberian Languages
Porto Salvo, Portugal
This paper reports our recent work on extending our previous gender detector, targeted only at distinguishing between adult male and female voices, to encompass children’s voices as well. The classifiers were based on multilayer perceptrons and Gaussian mixture models and used Perceptual Linear Prediction coefficients, plus deltas, and pitch as features. Despite the small amount of training data for children’s voices, fairly good results were obtained in a test corpus of similar recording conditions (minimum classification error rate of 2.6%). Tests on real life corpora revealed the expected degradation with noisy environments and distant microphones. Tests with transformed female voices intended as cartoon child characters showed that they were mostly classified as children’s voices.
Index Terms: gender detection, age effects, children voices
Bibliographic reference. Martins, Rui / Trancoso, Isabel / Abad, Alberto / Meinedo, Hugo (2009): "Detection of children²s voices", In SLTECH-2009, 77-80.