Multi-speaker/speaker-independent architectures for the multi-state time delay neural network

The authors present an improved multistate time delay neural network (MS-TDNN) for speaker-independent, connected letter recognition which outperforms an HMM (hidden Markov model) based system (SPHINX) and previous MS-TDNNs. They also explore new network architectures with internal speaker models. Four different architectures characterized by an increasing number of speaker-specific parameters are introduced. The speaker-specific parameters can be adjusted by automatic speaker identification or by speaker adaptation, allowing for tuning-in to a new speaker. Both methods lead to significant improvements over the straightforward speaker-independent architecture. Even unsupervised tuning-in (speech is unlabeled) works well.<
>