We present the Factorial Deep Markov Model (FDMM) for representation learning of speech. The FDMM learns disentangled, interpretable and lower dimensional latent representations from speech without supervision. We use a static and dynamic latent variable to exploit the fact that information in a speech signal evolves at different time scales. Latent representations learned by the FDMM outperform a baseline ivector system on speaker verification and dialect identification while also reducing the error rate of a phone recognition system in a domain mismatch scenario.

@inproceedings{sameer-et-al-icassp-19, abstract = {We present the Factorial Deep Markov Model (FDMM) for representation learning of speech. The FDMM learns disentangled, interpretable and lower dimensional latent representations from speech without supervision. We use a static and dynamic latent variable to exploit the fact that information in a speech signal evolves at different time scales. Latent representations learned by the FDMM outperform a baseline ivector system on speaker verification and dialect identification while also reducing the error rate of a phone recognition system in a domain mismatch scenario.}, address = {Brighton, UK}, author = {Sameer Khurana and Shafiq Joty and Ahmed Ali and James Glass}, booktitle = {International Conference on Acoustics, Speech, and Signal Processing}, month = {September}, pages = {6540 -- 6544}, publisher = {IEEE}, series = {ICASSP'19}, title = {A Fatorial Deep Markov Model For Unsupervised Disentangled Representation Learning From Speech}, url = {https://groups.csail.mit.edu/sls/publications/2019/SameerKhurana_ICASSP-2019.pdf}, year = {2019} }