dc.contributor.author | Kolesau, Aliaksei | |
dc.date.accessioned | 2023-09-18T20:34:42Z | |
dc.date.available | 2023-09-18T20:34:42Z | |
dc.date.issued | 2019 | |
dc.identifier.uri | https://etalpykla.vilniustech.lt/handle/123456789/151025 | |
dc.description.abstract | Many keyword spotting models use neural networks to detect acoustic events such as phonemes, word pieces or whole words. The model is inferenced on every frame (segmented piece of audio) which is typically every 10ms. In order to improve the quality of classification neural network uses audio features for both the frame under classification and several adjacent frames. This introduces a tradeoff. Too large receptive field might cause overfitting, increases the number of parameters and latency. Too small receptive field might not be able to provide enough information to correctly classify audio event. We investigate several policies of constructing receptive field for neural network in keyword spotting including the ways to make receptive field more sparse such as frame skipping and frame stacking. | eng |
dc.format | PDF | |
dc.format.extent | p. 39 | |
dc.format.medium | tekstas / txt | |
dc.language.iso | eng | |
dc.source.uri | https://www.zurnalai.vu.lt/proceedings/issue/view/1389 | |
dc.source.uri | https://doi.org/10.15388/Proceedings.2019.8 | |
dc.title | Receptive field in neural network keyword spotting models | |
dc.type | Konferencijos pranešimo santrauka / Conference presentation abstract | |
dcterms.accessRights | This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. | |
dcterms.references | 0 | |
dc.type.pubtype | T2 - Konferencijos pranešimo tezės / Conference presentation abstract | |
dc.contributor.institution | Vilniaus Gedimino technikos universitetas | |
dc.contributor.faculty | Fundamentinių mokslų fakultetas / Faculty of Fundamental Sciences | |
dc.subject.researchfield | T 007 - Informatikos inžinerija / Informatics engineering | |
dc.subject.studydirection | B04 - Informatikos inžinerija / Informatics engineering | |
dc.subject.vgtuprioritizedfields | IK0303 - Dirbtinio intelekto ir sprendimų priėmimo sistemos / Artificial intelligence and decision support systems | |
dc.subject.ltspecializations | L106 - Transportas, logistika ir informacinės ir ryšių technologijos (IRT) / Transport, logistic and information and communication technologies | |
dc.subject.en | keyword spotting models | |
dc.subject.en | neural networks | |
dc.subject.en | audio features | |
dcterms.sourcetitle | 11th international workshop on data analysis methods for software systems (DAMSS 2019), Druskininkai, Lithuania, November 28-30, 2019 / Lithuanian Computer Society, Vilnius University Institute of Data Science and Digital Technologies, Lithuanian Academy of Sciences | |
dc.publisher.name | Vilnius University | |
dc.publisher.city | 2019 | |
dc.identifier.doi | 10.15388/Proceedings.2019.8 | |
dc.identifier.elaba | 76603619 | |