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

Cued Speech (CS) is a visual coding tool to encode spoken languages at thephonetic level, which combines lip-reading and hand gestures to effectivelyassist communication among people with hearing impairments. The Automatic CSRecognition (ACSR) task aims to recognize CS videos into linguistic texts,which involves both lips and hands as two distinct modalities conveyingcomplementary information. However, the traditional centralized trainingapproach poses potential privacy risks due to the use of facial and gesturevideos in CS data. To address this issue, we propose a new Federated CuedSpeech Recognition (FedCSR) framework to train an ACSR model over thedecentralized CS data without sharing private information. In particular, amutual knowledge distillation method is proposed to maintain cross-modalsemantic consistency of the Non-IID CS data, which ensures learning a unifiedfeature space for both linguistic and visual information. On the server side, aglobally shared linguistic model is trained to capture the long-termdependencies in the text sentences, which is aligned with the visualinformation from the local clients via visual-to-linguistic distillation. Onthe client side, the visual model of each client is trained with its own localdata, assisted by linguistic-to-visual distillation treating the linguisticmodel as the teacher. To the best of our knowledge, this is the first approachto consider the federated ACSR task for privacy protection. Experimentalresults on the Chinese CS dataset with multiple cuers demonstrate that ourapproach outperforms both mainstream federated learning baselines and existingcentralized state-of-the-art ACSR methods, achieving 9.7% performanceimprovement for character error rate (CER) and 15.0% for word error rate (WER).

Cuing Without Sharing: A Federated Cued Speech Recognition Framework via Mutual Knowledge Distillation