Non-Parametric Bayesian Subspace Models for Acoustic Unit Discovery

Unsupervised learning, non- parametricBayesian models, acoustic unit discovery

This work investigates subspace non-parametric models for the task of learning a set of acoustic units fromunlabeled speech recordings. We constrain the base-measure of a Dirichlet- Process mixture with a phonetic subspaceestimated from other source languagesto build an educated prior, thereby forcing the learned acoustic units to resemble phones of known source languages. Two types of models are proposed: (i) the Subspace HMM (SHMM) which assumes that the phonetic subspace is the same for every language, (ii) the Hierarchical-Subspace HMM (H-SHMM) which relaxes this assumption and allows to have a languagespecific subspace estimated on the unlabeled target data. These models are applied on 3 languages: English, Yoruba and Mboshi and they are compared with various competitive acoustic units discovery baselines. Experimental results show that both subspace models outperform other systems in terms of clustering quality and segmentation accuracy. Moreover, we observe that the H-SHMM provides results superior to the SHMM supporting the idea that language-specific priors are preferable to language-agnostic priors for acoustic unit discovery.