• search hit 6 of 89
Back to Result List

State Sequence Pooling Training of Acoustic Models for Keyword Spotting

  • We propose a new training method to improve HMM-based keyword spotting. The loss function is based on a score computed with the keyword/filler model from the entire input sequence. It is equivalent to max/attention pooling but is based on prior acoustic knowledge. We also employ a multi-task learning setup by predicting both LVCSR and keyword posteriors. We compare our model to a baseline trained on frame-wise cross entropy, with and without per-class weighting. We employ a low-footprint TDNN for acoustic modeling. The proposed training yields significant and consistent improvement over the baseline in adverse noise conditions. The FRR on cafeteria noise is reduced from 13.07% to 5.28% at 9 dB SNR and from 37.44% to 6.78% at 5 dB SNR. We obtain these results with only 600 unique training keyword samples. The training method is independent of the frontend and acoustic model topology.

Export metadata

Additional Services

Search Google Scholar
Metadaten
Author:Kuba Lopatka, Tobias Bocklet
DOI:https://doi.org/10.21437/Interspeech.2020-2722
ISSN:2958-1796
Parent Title (English):Proceedings Interspeech 2020
Document Type:conference proceeding (article)
Language:English
Reviewed:Begutachtet/Reviewed
Release Date:2024/07/03
Tag:keyword spotting, machine learning, speech recognition
Pagenumber:5
First Page:4338
Last Page:4342
institutes:Fakultät Informatik
Licence (German):Keine Lizenz - Deutsches Urheberrecht gilt
Verstanden ✔
Diese Webseite verwendet technisch erforderliche Session-Cookies. Durch die weitere Nutzung der Webseite stimmen Sie diesem zu. Unsere Datenschutzerklärung finden Sie hier.