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

Speech separation remains an important area of multi-speaker signalprocessing. Deep neural network (DNN) models have attained the best performanceon many speech separation benchmarks. Some of these models can take significanttime to train and have high memory requirements. Previous work has proposedshortening training examples to address these issues but the impact of this onmodel performance is not yet well understood. In this work, the impact ofapplying these training signal length (TSL) limits is analysed for two speechseparation models: SepFormer, a transformer model, and Conv-TasNet, aconvolutional model. The WJS0-2Mix, WHAMR and Libri2Mix datasets are analysedin terms of signal length distribution and its impact on training efficiency.It is demonstrated that, for specific distributions, applying specific TSLlimits results in better performance. This is shown to be mainly due torandomly sampling the start index of the waveforms resulting in more uniqueexamples for training. A SepFormer model trained using a TSL limit of 4.42s anddynamic mixing (DM) is shown to match the best-performing SepFormer modeltrained with DM and unlimited signal lengths. Furthermore, the 4.42s TSL limitresults in a 44% reduction in training time with WHAMR.

On Data Sampling Strategies for Training Neural Network Speech Separation Models