000008295 001__ 8295
000008295 005__ 20240418124754.0
000008295 0247_ $$2DOI$$a10.6083/4b29b666b
000008295 037__ $$aIR
000008295 041__ $$aeng
000008295 245__ $$aUsing conditional adversarial networks for intelligibility improvement for dysarthric speech and laryngectomees
000008295 260__ $$bOregon Health and Science University
000008295 269__ $$a2020
000008295 336__ $$aAbstract
000008295 520__ $$aWe explored voice conversion systems to improve speech intelligibility of 1) dysarthric speech and 2) laryngectomees.
In the first case, we explore the potential of conditional generative adversarial networks (cGANs) to learn the mapping from habitual speech to clear speech. We evaluated the performance of cGANs in three tasks: 1) speaker-dependent one-to-one mappings, 2) speaker-independent many-to-one mappings, and 3) speaker-independent many-to-many mappings. In the first task, cGANs outperformed a traditional deep learning (DNN) mapping in term of average keyword recall accuracy and the number of speakers with improved intelligibility. In the second task, we showed that without clear speech, we can significantly improve intelligibility of the habitual speech of one of three speakers. In the third task which is the most challenging one, we improved the keyword recall accuracy for two of three speakers.
In the second case, we aim to improve speech of laryngectomees in term of intelligibility and naturalness. We predict the voicing and voicing degree for laryngectomees from speech spectra using a deep neural network. We use a logarithmically falling synthetic F0 for statement phrases. Spectra are converted to synthetic target spectra using a cGAN.
000008295 540__ $$fCC BY
000008295 542__ $$fIn copyright - joint owners
000008295 650__ $$aSpeech Intelligibility$$026231
000008295 650__ $$aSpeech Acoustics$$026227
000008295 650__ $$aDeep Learning$$012734
000008295 650__ $$aLaryngectomy$$021330
000008295 6531_ $$alaryngectomees
000008295 6531_ $$aconditional adversarial nets
000008295 6531_ $$adysarthric speech
000008295 6531_ $$avoice conversion
000008295 691__ $$aSchool of Medicine$$041369
000008295 692__ $$aDepartment of Computer Science and Electrical Engineering$$041404
000008295 7001_ $$aDinh, Tuan$$uOregon Health and Science University$$041354
000008295 7001_ $$aKain, Alexander$$uOregon Health and Science University$$041354
000008295 711__ $$aResearch Week$$uOregon Health and Science University$$d2020
000008295 8564_ $$9850b77f3-bdb1-4f9e-9ed9-eedb7b541bf8$$s43637$$uhttps://digitalcollections.ohsu.edu/record/8295/files/Tuan-Dinh.pdf
000008295 8564_ $$9ea4ee87d-3091-43b6-96dd-269c582468d0$$s795822$$uhttps://digitalcollections.ohsu.edu/record/8295/files/Dinh_Presentation.pdf
000008295 905__ $$a/rest/prod/4b/29/b6/66/4b29b666b
000008295 980__ $$aResearch Week