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

Modern neural-network-based speech processing systems are typically requiredto be robust against reverberation, and the training of such systems thus needsa large amount of reverberant data. During the training of the systems,on-the-fly simulation pipeline is nowadays preferred as it allows the model totrain on infinite number of data samples without pre-generating and saving themon harddisk. An RIR simulation method thus needs to not only generate morerealistic artificial room impulse response (RIR) filters, but also generatethem in a fast way to accelerate the training process. Existing RIR simulationtools have proven effective in a wide range of speech processing tasks andneural network architectures, but their usage in on-the-fly simulation pipelineremains questionable due to their computational complexity or the quality ofthe generated RIR filters. In this paper, we propose FRAM-RIR, a fast randomapproximation method of the widely-used image-source method (ISM), toefficiently generate realistic multi-channel RIR filters. FRAM-RIR bypasses theexplicit calculation of sound propagation paths in ISM-based algorithms byrandomly sampling the location and number of reflections of each virtual soundsource based on several heuristic assumptions, while still maintains accuratedirection-of-arrival (DOA) information of all sound sources. Visualization oforacle beampatterns and directional features shows that FRAM-RIR can generatemore realistic RIR filters than existing widely-used ISM-based tools, andexperiment results on multi-channel noisy speech separation and dereverberationtasks with a wide range of neural network architectures show that modelstrained with FRAM-RIR can also achieve on par or better performance on realRIRs compared to other RIR simulation tools with a significantly acceleratedtraining procedure. A Python implementation of FRAM-RIR is released.

Fast Random Approximation of Multi-channel Room Impulse Response