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/

Zendy Plus

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

Zendy Tools

Zendy Open

In this paper, we investigate a deep learning approach for speech denoisingthrough an efficient ensemble of specialist neural networks. By splitting upthe speech denoising task into non-overlapping subproblems and introducing aclassifier, we are able to improve denoising performance while also reducingcomputational complexity. More specifically, the proposed model incorporates agating network which assigns noisy speech signals to an appropriate specialistnetwork based on either speech degradation level or speaker gender. In ourexperiments, a baseline recurrent network is compared against an ensemble ofsimilarly-designed smaller recurrent networks regulated by the auxiliary gatingnetwork. Using stochastically generated batches from a large noisy speechcorpus, the proposed model learns to estimate a time-frequency masking matrixbased on the magnitude spectrogram of an input mixture signal. Both baselineand specialist networks are trained to estimate the ideal ratio mask, while thegating network is trained to perform subproblem classification. Our findingsdemonstrate that a fine-tuned ensemble network is able to exceed the speechdenoising capabilities of a generalist network, doing so with fewer modelparameters.

Sparse Mixture of Local Experts for Efficient Speech Enhancement