Bayesian Subspace HMM for the Zerospeech 2020 Challenge

2020/05/19 by Bolaji Yusuf, Yusuf, Bolaji, Lucas Ondel +1
Computer Science · Engineering · Mathematics · #Audio and Speech Processing (eess.AS) #Computation and Language (cs.CL) #FOS: Computer and information sciences #FOS: Electrical engineering #Machine Learning (cs.LG) #Machine Learning (stat.ML) #cs.CL #cs.LG #eess.AS #electronic engineering #information engineering #stat.ML

paper · pdf · doi:10.48550/arxiv.2005.09282

arxiv created 2020/07/27 · arxiv updated 2020/07/28

Abstract

In this paper we describe our submission to the Zerospeech 2020 challenge, where the participants are required to discover latent representations from unannotated speech, and to use those representations to perform speech synthesis, with synthesis quality used as a proxy metric for the unit quality. In our system, we use the Bayesian Subspace Hidden Markov Model (SHMM) for unit discovery. The SHMM models each unit as an HMM whose parameters are constrained to lie in a low dimensional subspace of the total parameter space which is trained to model phonetic variability. Our system compares favorably with the baseline on the human-evaluated character error rate while maintaining significantly lower unit bitrate.

Related

• Bayesian Subspace Hidden Markov Model for Acoustic Unit Discovery
• Bayesian Models for Unit Discovery on a Very Low Resource Language
• An elitist approach for extracting automatically well-realized speech sounds with high confidence
• HMM-based data augmentation for E2E systems for building conversational speech synthesis systems
• HAM-TTS: Hierarchical Acoustic Modeling for Token-Based Zero-Shot Text-to-Speech with Model and Data Scaling