Using speaker group dependent modelling to improve fusion of fragmentary classifier decisions | IEEE Conference Publication | IEEE Xplore