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 is very important in real-world applications such ashuman-machine interaction, hearing aids devices, and automatic meetingtranscription. In recent years, a significant improvement occurred towards thesolution based on deep learning. In fact, much attention has been drawn tosupervised learning methods using synthetic mixtures datasets despite theirbeing not representative of real-world mixtures. The difficulty in building arealistic dataset led researchers to use unsupervised learning methods, becauseof their ability to handle realistic mixtures directly. The results ofunsupervised learning methods are still unconvincing. In this paper, a methodis introduced to create a realistic dataset with ground truth sources forspeech separation. The main challenge in designing a realistic dataset is theunavailability of ground truths for speakers signals. To address this, wepropose a method for simultaneously recording two speakers and obtaining theground truth for each. We present a methodology for benchmarking our realisticdataset using a deep learning model based on Bidirectional Gated RecurrentUnits (BGRU) and clustering algorithm. The experiments show that our proposeddataset improved SI-SDR (Scale Invariant Signal to Distortion Ratio) by 1.65 dBand PESQ (Perceptual Evaluation of Speech Quality) by approximately 0.5. Wealso evaluated the effectiveness of our method at different distances betweenthe microphone and the speakers and found that it improved the stability of thelearned model.

Towards Solving Cocktail-Party: The First Method to Build a Realistic Dataset with Ground Truths for Speech Separation