Bayesian Learning For Deep Neural Network Adaptation

Abstract

A key task for speech recognition systems is to reduce the mismatch between training and evaluation data that is often attributable to speaker differences. Speaker adaptation techniques play a vital role to reduce the mismatch. Model-based speaker adaptation approaches often require sufficient amounts of target speaker data to ensure robustness. When the amount of speaker level data is limited, speaker adaptation is prone to overfitting and poor generalization. To address the issue, this paper proposes a full Bayesian learning based DNN speaker adaptation framework to model speaker-dependent (SD) parameter uncertainty given limited speaker specific adaptation data. This framework is investigated in three forms of model based DNN adaptation techniques: Bayesian learning of hidden unit contributions (BLHUC), Bayesian parameterized activation functions (BPAct), and Bayesian hidden unit bias vectors (BHUB). In the three methods, deterministic SD parameters are replaced by latent variable p

Stats

• citations19
• S2 citations—
• github stars0
• HF likes0
• heat score9.76
• arxiv keyxie2020bayesian

Related papers

• Unsupervised Model-based Speaker Adaptation Of End-to-end Lattice-free MMI Model For Speech Recognition (2022)2.26
• Empirical Evaluation Of Speaker Adaptation On DNN Based Acoustic Model (2018)5.24
• Learning To Adapt: A Meta-learning Approach For Speaker Adaptation (2018)9.76
• Learning Hidden Unit Contributions For Unsupervised Acoustic Model Adaptation (2016)14.47
• Adversarial Speaker Adaptation (2019)10.21
• Generalized Domain Adaptation Framework For Parametric Back-end In Speaker Recognition (2023)0.00
• Confidence Score Based Speaker Adaptation Of Conformer Speech Recognition Systems (2023)8.35
• A Unified Speaker Adaptation Method For Speech Synthesis Using Transcribed And Untranscribed Speech With Backpropagation (2019)0.00