Matches in SemOpenAlex for { <https://semopenalex.org/work/W3095820034> ?p ?o ?g. }
Showing items 1 to 95 of
95
with 100 items per page.
- W3095820034 endingPage "96" @default.
- W3095820034 startingPage "80" @default.
- W3095820034 abstract "Accurately modelling the long-term dependencies of noisy speech is critical to the performance of a speech enhancement system. Current deep learning approaches to speech enhancement employ either a recurrent neural network (RNN) or a temporal convolutional network (TCN). However, RNNs and TCNs both demonstrate deficiencies when modelling long-term dependencies. Enter multi-head attention (MHA) — a mechanism that has outperformed both RNNs and TCNs in tasks such as machine translation. By using sequence similarity, MHA possesses the ability to more efficiently model long-term dependencies. Moreover, masking can be employed to ensure that the MHA mechanism remains causal — an attribute critical for real-time processing. Motivated by these points, we investigate a deep neural network (DNN) that utilises masked MHA for causal speech enhancement. The conditions used to evaluate the proposed DNN include real-world non-stationary and coloured noise sources at multiple SNR levels. Our extensive experimental investigation demonstrates that the proposed DNN can produce enhanced speech at a higher quality and intelligibility than both RNNs and TCNs. We conclude that deep learning approaches employing masked MHA are more suited for causal speech enhancement than RNNs and TCNs. Availability—MHANet is available at https://github.com/anicolson/DeepXi" @default.
- W3095820034 created "2020-11-09" @default.
- W3095820034 creator A5041494929 @default.
- W3095820034 creator A5046089538 @default.
- W3095820034 date "2020-12-01" @default.
- W3095820034 modified "2023-10-18" @default.
- W3095820034 title "Masked multi-head self-attention for causal speech enhancement" @default.
- W3095820034 cites W1998550344 @default.
- W3095820034 cites W2020997493 @default.
- W3095820034 cites W2044893557 @default.
- W3095820034 cites W2051428568 @default.
- W3095820034 cites W2069681747 @default.
- W3095820034 cites W2078528584 @default.
- W3095820034 cites W2107878631 @default.
- W3095820034 cites W2115144768 @default.
- W3095820034 cites W2121973264 @default.
- W3095820034 cites W2131774270 @default.
- W3095820034 cites W2138196899 @default.
- W3095820034 cites W2141998673 @default.
- W3095820034 cites W2144404214 @default.
- W3095820034 cites W2147817981 @default.
- W3095820034 cites W2159202424 @default.
- W3095820034 cites W2678916739 @default.
- W3095820034 cites W2789260495 @default.
- W3095820034 cites W2805233667 @default.
- W3095820034 cites W2946615417 @default.
- W3095820034 cites W2952979007 @default.
- W3095820034 cites W2972592847 @default.
- W3095820034 cites W2978520562 @default.
- W3095820034 cites W3005262054 @default.
- W3095820034 cites W3017350693 @default.
- W3095820034 cites W3031404175 @default.
- W3095820034 cites W3092531370 @default.
- W3095820034 cites W3105888187 @default.
- W3095820034 cites W3147539069 @default.
- W3095820034 cites W3159776099 @default.
- W3095820034 cites W4232282348 @default.
- W3095820034 cites W4250584829 @default.
- W3095820034 doi "https://doi.org/10.1016/j.specom.2020.10.004" @default.
- W3095820034 hasPublicationYear "2020" @default.
- W3095820034 type Work @default.
- W3095820034 sameAs 3095820034 @default.
- W3095820034 citedByCount "40" @default.
- W3095820034 countsByYear W30958200342021 @default.
- W3095820034 countsByYear W30958200342022 @default.
- W3095820034 countsByYear W30958200342023 @default.
- W3095820034 crossrefType "journal-article" @default.
- W3095820034 hasAuthorship W3095820034A5041494929 @default.
- W3095820034 hasAuthorship W3095820034A5046089538 @default.
- W3095820034 hasConcept C108583219 @default.
- W3095820034 hasConcept C111472728 @default.
- W3095820034 hasConcept C138885662 @default.
- W3095820034 hasConcept C147168706 @default.
- W3095820034 hasConcept C154945302 @default.
- W3095820034 hasConcept C163294075 @default.
- W3095820034 hasConcept C2776182073 @default.
- W3095820034 hasConcept C28490314 @default.
- W3095820034 hasConcept C41008148 @default.
- W3095820034 hasConcept C50644808 @default.
- W3095820034 hasConcept C60048801 @default.
- W3095820034 hasConcept C61328038 @default.
- W3095820034 hasConcept C81363708 @default.
- W3095820034 hasConceptScore W3095820034C108583219 @default.
- W3095820034 hasConceptScore W3095820034C111472728 @default.
- W3095820034 hasConceptScore W3095820034C138885662 @default.
- W3095820034 hasConceptScore W3095820034C147168706 @default.
- W3095820034 hasConceptScore W3095820034C154945302 @default.
- W3095820034 hasConceptScore W3095820034C163294075 @default.
- W3095820034 hasConceptScore W3095820034C2776182073 @default.
- W3095820034 hasConceptScore W3095820034C28490314 @default.
- W3095820034 hasConceptScore W3095820034C41008148 @default.
- W3095820034 hasConceptScore W3095820034C50644808 @default.
- W3095820034 hasConceptScore W3095820034C60048801 @default.
- W3095820034 hasConceptScore W3095820034C61328038 @default.
- W3095820034 hasConceptScore W3095820034C81363708 @default.
- W3095820034 hasLocation W30958200341 @default.
- W3095820034 hasOpenAccess W3095820034 @default.
- W3095820034 hasPrimaryLocation W30958200341 @default.
- W3095820034 hasRelatedWork W1986772939 @default.
- W3095820034 hasRelatedWork W2037635165 @default.
- W3095820034 hasRelatedWork W2139175112 @default.
- W3095820034 hasRelatedWork W2140410589 @default.
- W3095820034 hasRelatedWork W2144314030 @default.
- W3095820034 hasRelatedWork W2542098180 @default.
- W3095820034 hasRelatedWork W2891283181 @default.
- W3095820034 hasRelatedWork W3129072390 @default.
- W3095820034 hasRelatedWork W4200562864 @default.
- W3095820034 hasRelatedWork W4375869276 @default.
- W3095820034 hasVolume "125" @default.
- W3095820034 isParatext "false" @default.
- W3095820034 isRetracted "false" @default.
- W3095820034 magId "3095820034" @default.
- W3095820034 workType "article" @default.