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

This paper proposes a new paradigm for handling far-field multi-speaker datain an end-to-end neural network manner, called directional automatic speechrecognition (D-ASR), which explicitly models source speaker locations. InD-ASR, the azimuth angle of the sources with respect to the microphone array isdefined as a latent variable. This angle controls the quality of separation,which in turn determines the ASR performance. All three functionalities ofD-ASR: localization, separation, and recognition are connected as a singledifferentiable neural network and trained solely based on ASR errorminimization objectives. The advantages of D-ASR over existing methods arethreefold: (1) it provides explicit speaker locations, (2) it improves theexplainability factor, and (3) it achieves better ASR performance as theprocess is more streamlined. In addition, D-ASR does not require explicitdirection of arrival (DOA) supervision like existing data-driven localizationmodels, which makes it more appropriate for realistic data. For the case of twosource mixtures, D-ASR achieves an average DOA prediction error of less thanthree degrees. It also outperforms a strong far-field multi-speaker end-to-endsystem in both separation quality and ASR performance.

Directional ASR: A New Paradigm for E2E Multi-Speaker Speech Recognition with Source Localization