Matches in SemOpenAlex for { <https://semopenalex.org/work/W2892365986> ?p ?o ?g. }
- W2892365986 abstract "Robust speech processing in multitalker acoustic environments requiresautomatic speech separation. While single-channel, speaker-independent speechseparation methods have recently seen great progress, the accuracy, latency,and computational cost of speech separation remain insufficient. The majorityof the previous methods have formulated the separation problem through thetime-frequency representation of the mixed signal, which has several drawbacks,including the decoupling of the phase and magnitude of the signal, thesuboptimality of spectrogram representations for speech separation, and thelong latency in calculating the spectrogram. To address these shortcomings, wepropose the time-domain audio separation network (TasNet), which is a deeplearning autoencoder framework for time-domain speech separation. TasNet uses aconvolutional encoder to create a representation of the signal that isoptimized for extracting individual speakers. Speaker extraction is achieved byapplying a weighting function (mask) to the encoder output. The modifiedencoder representation is then inverted to the sound waveform using a lineardecoder. The masks are found using a temporal convolutional network consistingof dilated convolutions, which allow the network to model the long-termdependencies of the speech signal. This end-to-end speech separation algorithmsignificantly outperforms previous time-frequency methods in terms ofseparating speakers in mixed audio, even when compared to the separationaccuracy achieved with the ideal time-frequency mask of the speakers. Inaddition, TasNet has a smaller model size and a shorter minimum latency, makingit a suitable solution for both offline and real-time speech separationapplications. This study therefore represents a major step toward actualizingspeech separation for real-world speech processing technologies." @default.
- W2892365986 created "2018-09-27" @default.
- W2892365986 creator A5033351155 @default.
- W2892365986 creator A5048439332 @default.
- W2892365986 date "2018-09-20" @default.
- W2892365986 modified "2023-10-17" @default.
- W2892365986 title "TasNet: Surpassing Ideal Time-Frequency Masking for Speech Separation." @default.
- W2892365986 cites W1482149378 @default.
- W2892365986 cites W1485161427 @default.
- W2892365986 cites W1528954144 @default.
- W2892365986 cites W153881393 @default.
- W2892365986 cites W1552314771 @default.
- W2892365986 cites W1677182931 @default.
- W2892365986 cites W1845880232 @default.
- W2892365986 cites W1901129140 @default.
- W2892365986 cites W1980051803 @default.
- W2892365986 cites W1994514225 @default.
- W2892365986 cites W1994923416 @default.
- W2892365986 cites W2039057510 @default.
- W2892365986 cites W2044893557 @default.
- W2892365986 cites W2069681747 @default.
- W2892365986 cites W2078528584 @default.
- W2892365986 cites W2079362249 @default.
- W2892365986 cites W2079724265 @default.
- W2892365986 cites W2096980176 @default.
- W2892365986 cites W2139896607 @default.
- W2892365986 cites W2144763279 @default.
- W2892365986 cites W2147455188 @default.
- W2892365986 cites W2161219071 @default.
- W2892365986 cites W2221409856 @default.
- W2892365986 cites W2401387233 @default.
- W2892365986 cites W2403380333 @default.
- W2892365986 cites W2405774341 @default.
- W2892365986 cites W2550143307 @default.
- W2892365986 cites W2552071709 @default.
- W2892365986 cites W2558649592 @default.
- W2892365986 cites W2568308529 @default.
- W2892365986 cites W2612445135 @default.
- W2892365986 cites W2734774145 @default.
- W2892365986 cites W2774707525 @default.
- W2892365986 cites W2792764867 @default.
- W2892365986 cites W2800022361 @default.
- W2892365986 cites W285277413 @default.
- W2892365986 cites W2889540509 @default.
- W2892365986 cites W2891405874 @default.
- W2892365986 cites W2892163332 @default.
- W2892365986 cites W2949117887 @default.
- W2892365986 cites W2953333557 @default.
- W2892365986 cites W2962866211 @default.
- W2892365986 cites W2962935966 @default.
- W2892365986 cites W2963045393 @default.
- W2892365986 cites W2963163009 @default.
- W2892365986 cites W2963301902 @default.
- W2892365986 hasPublicationYear "2018" @default.
- W2892365986 type Work @default.
- W2892365986 sameAs 2892365986 @default.
- W2892365986 citedByCount "24" @default.
- W2892365986 countsByYear W28923659862018 @default.
- W2892365986 countsByYear W28923659862019 @default.
- W2892365986 countsByYear W28923659862020 @default.
- W2892365986 countsByYear W28923659862021 @default.
- W2892365986 crossrefType "posted-content" @default.
- W2892365986 hasAuthorship W2892365986A5033351155 @default.
- W2892365986 hasAuthorship W2892365986A5048439332 @default.
- W2892365986 hasConcept C101738243 @default.
- W2892365986 hasConcept C103824480 @default.
- W2892365986 hasConcept C108583219 @default.
- W2892365986 hasConcept C111919701 @default.
- W2892365986 hasConcept C118505674 @default.
- W2892365986 hasConcept C153180895 @default.
- W2892365986 hasConcept C154945302 @default.
- W2892365986 hasConcept C19118579 @default.
- W2892365986 hasConcept C2776864781 @default.
- W2892365986 hasConcept C28490314 @default.
- W2892365986 hasConcept C31972630 @default.
- W2892365986 hasConcept C41008148 @default.
- W2892365986 hasConcept C45273575 @default.
- W2892365986 hasConcept C61328038 @default.
- W2892365986 hasConceptScore W2892365986C101738243 @default.
- W2892365986 hasConceptScore W2892365986C103824480 @default.
- W2892365986 hasConceptScore W2892365986C108583219 @default.
- W2892365986 hasConceptScore W2892365986C111919701 @default.
- W2892365986 hasConceptScore W2892365986C118505674 @default.
- W2892365986 hasConceptScore W2892365986C153180895 @default.
- W2892365986 hasConceptScore W2892365986C154945302 @default.
- W2892365986 hasConceptScore W2892365986C19118579 @default.
- W2892365986 hasConceptScore W2892365986C2776864781 @default.
- W2892365986 hasConceptScore W2892365986C28490314 @default.
- W2892365986 hasConceptScore W2892365986C31972630 @default.
- W2892365986 hasConceptScore W2892365986C41008148 @default.
- W2892365986 hasConceptScore W2892365986C45273575 @default.
- W2892365986 hasConceptScore W2892365986C61328038 @default.
- W2892365986 hasLocation W28923659861 @default.
- W2892365986 hasOpenAccess W2892365986 @default.
- W2892365986 hasPrimaryLocation W28923659861 @default.
- W2892365986 hasRelatedWork W1482149378 @default.
- W2892365986 hasRelatedWork W1552314771 @default.
- W2892365986 hasRelatedWork W2069681747 @default.
- W2892365986 hasRelatedWork W2127851351 @default.
- W2892365986 hasRelatedWork W2141998673 @default.