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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

The joint training framework for speech enhancement and recognition methodshave obtained quite good performances for robust end-to-end automatic speechrecognition (ASR). However, these methods only utilize the enhanced feature asthe input of the speech recognition component, which are affected by the speechdistortion problem. In order to address this problem, this paper proposes agated recurrent fusion (GRF) method with joint training framework for robustend-to-end ASR. The GRF algorithm is used to dynamically combine the noisy andenhanced features. Therefore, the GRF can not only remove the noise signalsfrom the enhanced features, but also learn the raw fine structures from thenoisy features so that it can alleviate the speech distortion. The proposedmethod consists of speech enhancement, GRF and speech recognition. Firstly, themask based speech enhancement network is applied to enhance the input speech.Secondly, the GRF is applied to address the speech distortion problem. Thirdly,to improve the performance of ASR, the state-of-the-art speech transformeralgorithm is used as the speech recognition component. Finally, the jointtraining framework is utilized to optimize these three components,simultaneously. Our experiments are conducted on an open-source Mandarin speechcorpus called AISHELL-1. Experimental results show that the proposed methodachieves the relative character error rate (CER) reduction of 10.04\% over theconventional joint enhancement and transformer method only using the enhancedfeatures. Especially for the low signal-to-noise ratio (0 dB), our proposedmethod can achieves better performances with 12.67\% CER reduction, whichsuggests the potential of our proposed method.

Gated Recurrent Fusion with Joint Training Framework for Robust End-to-End Speech Recognition