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 Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

The acoustic variability of noisy and reverberant speech mixtures isinfluenced by multiple factors, such as the spectro-temporal characteristics ofthe target speaker and the interfering noise, the signal-to-noise ratio (SNR)and the room characteristics. This large variability poses a major challengefor learning-based speech enhancement systems, since a mismatch between thetraining and testing conditions can substantially reduce the performance of thesystem. Generalization to unseen conditions is typically assessed by testingthe system with a new speech, noise or binaural room impulse response (BRIR)database different from the one used during training. However, the difficultyof the speech enhancement task can change across databases, which cansubstantially influence the results. The present study introduces ageneralization assessment framework that uses a reference model trained on thetest condition, such that it can be used as a proxy for the difficulty of thetest condition. This allows to disentangle the effect of the change in taskdifficulty from the effect of dealing with new data, and thus to define a newmeasure of generalization performance termed the generalization gap. Theprocedure is repeated in a cross-validation fashion by cycling through multiplespeech, noise, and BRIR databases to accurately estimate the generalizationgap. The proposed framework is applied to evaluate the generalization potentialof a feedforward neural network (FFNN), Conv-TasNet, DCCRN and MANNER. We findthat for all models, the performance degrades the most in speech mismatches,while good noise and room generalization can be achieved by training onmultiple databases. Moreover, while recent models show higher performance inmatched conditions, their performance substantially decreases in mismatchedconditions and can become inferior to that of the FFNN-based system.

Assessing the Generalization Gap of Learning-Based Speech Enhancement Systems in Noisy and Reverberant Environments