English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Dysarthric speech recognition is a challenging task as dysarthric data islimited and its acoustics deviate significantly from normal speech. Model-basedspeaker adaptation is a promising method by using the limited dysarthric speechto fine-tune a base model that has been pre-trained from large amounts ofnormal speech to obtain speaker-dependent models. However, statisticdistribution mismatches between the normal and dysarthric speech data limit theadaptation performance of the base model. To address this problem, we proposeto re-initialize the base model via meta-learning to obtain a better modelinitialization. Specifically, we focus on end-to-end models and extend themodel-agnostic meta learning (MAML) and Reptile algorithms to meta update thebase model by repeatedly simulating adaptation to different dysarthricspeakers. As a result, the re-initialized model acquires dysarthric speechknowledge and learns how to perform fast adaptation to unseen dysarthricspeakers with improved performance. Experimental results on UASpeech datasetshow that the best model with proposed methods achieves 54.2% and 7.6% relativeword error rate reduction compared with the base model without finetuning andthe model directly fine-tuned from the base model, respectively, and it iscomparable with the state-of-the-art hybrid DNN-HMM model.

Improved End-to-End Dysarthric Speech Recognition via Meta-learning Based Model Re-initialization