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

Mean square error (MSE) has been the preferred choice as loss function in thecurrent deep neural network (DNN) based speech separation techniques. In thispaper, we propose a new cost function with the aim of optimizing the extendedshort time objective intelligibility (ESTOI) measure. We focus on applicationswhere low algorithmic latency ($\leq 10$ ms) is important. We use longshort-term memory networks (LSTM) and evaluate our proposed approach on foursets of two-speaker mixtures from extended Danish hearing in noise (HINT)dataset. We show that the proposed loss function can offer improved or at parobjective intelligibility (in terms of ESTOI) compared to an MSE optimizedbaseline while resulting in lower objective separation performance (in terms ofthe source to distortion ratio (SDR)). We then proceed to propose an approachwhere the network is first initialized with weights optimized for MSE criterionand then trained with the proposed ESTOI loss criterion. This approachmitigates some of the losses in objective separation performance whilepreserving the gains in objective intelligibility.

Deep neural network based speech separation optimizing an objective estimator of intelligibility for low latency applications