Variational Inference for Acoustic Unit Discovery

In this article we proposed to train a nonparametric Bayesian model for automatic units discovery within the Variational Bayesframework. Besides simplifying the training scheme, this approach proves to be fast and yields better solution whichmakes it more suitable for big databases. However, despite the improvement observed, the model still have difficultieswith the diversity of speech and tends to learn a large part of unwanted variability. The HMM model for speechsegment is convenient but unrealistic and most likely, stronger model will be needed if one wants to achieve accurate automatic units discovery. We plan to extent the present work by using the VB inference with more complex models, as in13, and to gain leverage of Bayesian language models14 to further improve the accuracy of the discovered units.