Multiple speaker tracking and detection:: Handset normalization and duration scoring

We describe SRI's speaker tracking and detection system in the NIST 1998 Speaker Detection and Tracking Development Evaluation. The system is designed for tracking switchboard conversations and uses a two-speaker and silence hidden Markov model (HMM) with a minimum state duration constraint and Gaussian mixture model (GMM) state distributions adapted from a single gender- and handset-independent imposter model distribution. Speaker tracking is used to segment waveforms for speaker detection, which is carried out by averaging frame scores of the Viterbi path and normalizing for handset variation via a novel parameter interpolation extension of HNORM for use with waveform segments of arbitrary lengths. A short-duration penalty to augment the acoustic scores is also introduced via a nonlinear combination function. Results on the NIST 1998 Speaker Detection and Tracking Development Evaluation dataset are reported, (C) 2000 Academic Press.