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

This paper proposes a deep speech enhancement method which exploits the highpotential of residual connections in a wide neural network architecture, atopology known as Wide Residual Network. This is supported on singledimensional convolutions computed alongside the time domain, which is apowerful approach to process contextually correlated representations throughthe temporal domain, such as speech feature sequences. We find the residualmechanism extremely useful for the enhancement task since the signal always hasa linear shortcut and the non-linear path enhances it in several steps byadding or subtracting corrections. The enhancement capacity of the proposal isassessed by objective quality metrics and the performance of a speechrecognition system. This was evaluated in the framework of the REVERB Challengedataset, including simulated and real samples of reverberated and noisy speechsignals. Results showed that enhanced speech from the proposed method succeededfor both, the enhancement task with intelligibility purposes and the speechrecognition system. The DNN model, trained with artificial synthesizedreverberation data, was able to deal with far-field reverberated speech fromreal scenarios. Furthermore, the method was able to take advantage of theresidual connection achieving to enhance signals with low noise level, which isusually a strong handicap of traditional enhancement methods.

Deep Speech Enhancement for Reverberated and Noisy Signals using Wide Residual Networks