The 1st Workshop on Child, Computer and Interaction (WOCCI2008)

Chania, Crete, Greece
October 23, 2008

A Comparison of Read and Spontaneous Children’s Speech Recognition

Matteo Gerosa, Diego Giuliani

FBK Fondazione Bruno Kessler, Povo (TN), Italy

This paper presents comparative analyses, and recognition experiments, on read and spontaneous Italian speech collected from children. The presented analyses focus on linguistic variations, variations in phone duration, and the scattering of phones in the acoustic space. The aim of these analyses is to achieve a better understanding of acoustic and linguistic difference between read and spontaneous speech uttered by children in the same age range (9-11). A recognition system was developed exploiting clean read speech, collected from children aged 7-13, and written texts. Results of phone and word recognition experiments, carried out with this system on read and spontaneous speech, are presented. Results of recognition experiments show that very high recognition performance can be achieved on clean read children’s speech (6.9% phone error rare). However, performance drops drastically when the system is applied to spontaneous speech collected from children (27.2% phone error rate).

Full Paper

Bibliographic reference.  Gerosa, Matteo / Giuliani, Diego (2008): "A comparison of read and spontaneous children²s speech recognition", In WOCCI-2008, paper 05.