Direct training of subspace distribution clustering hidden Markov model

It generally takes a long time and requires a large amount of speech data to train hidden Markov models for a speech recognition task of a reasonably large vocabulary. Recently, we proposed a compact acoustic model called “subspace distribution clustering hidden Markov model” (SDCHMM) with an aim to save some of the training effort. SDCHMMs are derived from tying continuous density hidden Markov models (CDHMMs) at a finer subphonetic level, namely the subspace distributions. Experiments on the Airline Travel Information System (ATIS) task show that SDCHMMs with significantly fewer model parameters—by one to two orders of magnitude—can be converted from CDHMMs with no loss in word accuracy [1], [2]. With such compact acoustic models, one should be able to train SDCHMMs directly from significantly less speech data (without intermediate CDHMMs). In this paper, we devise a direct SDCHMM training algorithm, assuming an a priori knowledge of the subspace distribution tying structure. On the ATIS task, it is found that both a context-independent and a context-dependent speaker-independent 20-stream SDCHMM system trained with 8 min of speech perform as well as their corresponding CDHMM system trained with 105 min and 36 h of speech, respectively.