Ubuyobozi bwa Audio AI

Conv-TasNet Igihe-Gutandukana

Conv-TasNet numuyoboro udasanzwe utandukanya amajwi avanze (nkabantu babiri bavugira icyarimwe) mukora muburyo butaziguye kumajwi mbisi aho gukora spekrogramu.

2 min somaIbiherutse kuvugururwa

Incamake

It matters because it set a new bar for speech separation quality while running fast enough for real-time use.

Kwibira cyane

Sisitemu yo gutandukanya gakondo ihindura amajwi kuri spekrogramu, gutandukanya imirongo, hanyuma uhindure inyuma, itakaza amakuru yicyiciro kandi caps ubuziranenge. Conv-TasNet (2019, Luo na Mesgarani) isimbuka rwose. Ikoresha kodegisi yize (1D imyanzuro) kugirango ihindure uduce duto duto twerekana imiterere ihindagurika yimbere, umuyoboro wo gutandukanya ugereranya mask kuri buri muvugizi, hamwe na decoder yize yubaka buri cyerekezo gisukuye. Gutandukanya ni igipande cyagutse cya 1D cyiswe Umuyoboro wigihe gito (TCN), gifata intera ndende ntisubire. Yatojwe nigipimo-kidahinduka SI-SNR igihombo hamwe namahugurwa ya permis-invariant, yarenze masike ya spekrogramu nziza, igisubizo cyigeze gutekerezwa ko ari umupaka wo hejuru.

Ubushishozi

Amayeri yibanze ni ugusimbuza Ihinduka rigufi-Igihe cya Fourier Guhindura hamwe na kodegisi yize ya 1D-yemeza, bityo umuyoboro ugasanga amajwi yerekana neza neza kugirango ahishe aho kuba agenewe kureba abantu. Itandukanyirizo rya TCN rikoresha impagarike yagutse hamwe nibintu bigenda byiyongera bikabije, bigatanga umurima munini wakira mugihe ugumye ugereranije. Masike igwiza kodegisi iranga ibintu-bifite ubwenge, kandi imyanzuro ihindagurika ihindura buri maskike igaruka kumurongo.

Ingaruka z'Ingamba

Kugera no kugera

Itezimbere kugerwaho binyuze mu kwandukura, kuvuga, no guhuza amajwi.

Igiciro na bije

Amatsinda yibitangazamakuru arashobora kohereza amajwi yihuse hamwe na bije nto.

Umuvuduko n'igipimo

Sisitemu ireba abakiriya irashobora gutunganya imikoranire ivugwa murwego runini.

Kazoza ka Conv-TasNet Igihe-Gutandukana

Conv-TasNet yabibye umuryango wose wigihe-indangarugero. Abasimbuye nka DPRNN, SepFormer, na TF-GridNet basunikishije ubuziranenge bwo gutandukana cyane, ariko Conv-TasNet ikomeje kuba urwego rukomeye, rworoheje kandi ruracyashyirwa mubikoresho aho compte iba ikomeye. Tegereza igishushanyo mbonera cya TCN kugirango ukomeze kugaragara mubikoresho bifasha kumva, gutwi, hamwe ninama nyayo-nyayo, akenshi itandukanijwe cyangwa igereranijwe kugirango ikore muri milisegonda kuri chip zigendanwa.

Gushyira mu bikorwa Isi

Gutandukanya abavuga rikijyana mu nama yafashwe amajwi kugirango buriwese yandukurwe neza.

Gutezimbere imvugo mumatwi hamwe nibikoresho bifasha kumva bitandukanya abavuga intego kubiganiro byinyuma.

Mbere yo gutunganya urusaku rwitwa-center amajwi mbere yo kuyigaburira kumenyekanisha imvugo.

Gusukura ibiganiro byuzuye muri podcast cyangwa firime nyuma yo gukora.

Ingaruka & Kurinda

Gukoresha nabi amajwi no kwigira ibyago byiyongera mugihe uruhushya rubuze.

Ukuri kurashobora kugabanuka hejuru yimvugo, imvugo, cyangwa urusaku rwibidukikije.

Amajwi yubukorikori arashobora kwibeshya kumvugo yukuri nta kirango gisobanutse.

Igishushanyo mbonera

1

Shaka uruhushya rusobanutse rwo gufata amajwi, gukoroniza, no gukoresha.

2

Ikizamini cyiza mubiganiro bitandukanye hamwe nuburyo bwimbere.

3

Sobanura igihe umuntu agomba gusuzuma cyangwa kwemeza ibisubizo.

4

Andika amajwi yubukorikori kandi ugumane inyandiko zerekana kubazwa.

Komeza Ubushakashatsi

Free newsletter

Get the daily AI briefing

Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.

One email each weekday. Unsubscribe in one click. We never sell or share your address.

Test yourself

Take the Conv-TasNet Time-Domain Separation quiz

Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.

Tangira ikibazo

Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation

Ubuyobozi bukurikira

Gufungura-Unmix Gutandukanya Umuziki

Ibibazo bikunze kubazwa

What is Conv-TasNet Time-Domain Separation?

Conv-TasNet numuyoboro udasanzwe utandukanya amajwi avanze (nkabantu babiri bavugira icyarimwe) mukora muburyo butaziguye kumajwi mbisi aho gukora spekrogramu. Nibyingenzi kuko yashyizeho umurongo mushya wo gutandukanya imvugo mugihe wiruka byihuse kugirango ukoreshe igihe.

Ni ubuhe buryo busobanura Conv-TasNet ugereranije na sisitemu yo gutandukana mbere?

Conv-TasNet isimbuza umuyoboro uhamye wa STFT hamwe na encoder / decoder wize bityo itandukanya amajwi mugihe cyagenwe kumurongo wibanze.

Ni ubuhe bwoko bw'urusobe Conv-TasNet ikoresha nk'uburyo bwo gutandukana?

Gutandukanya ni TCN yubatswe kuva yegeranye 1D ihindagurika, itanga umurima munini wakira mugihe usigaye ugereranije.

Niki gisimbuza impinduka gakondo-Igihe gito cya Fourier muri Conv-TasNet?

Aho kugirango STFT ihamye, yize 1D imyanzuro ikubiyemo uduce twinshi muburyo bwo kwerekana neza.

Nibihe bikorwa byo gutakaza nibyingenzi mumahugurwa Conv-TasNet?

Conv-TasNet yatojwe hamwe na SI-SNR igihombo hamwe na permisation-invariant imyitozo kugirango ikemure gahunda itazwi yabavuga batandukanye.

Ni ibihe bisubizo bigaragara Conv-TasNet yagezeho mu gutandukanya imvugo?

Mu kwirinda icyiciro cyo gutakaza uburyo bwa spekrogramu, Conv-TasNet yarenze igihe cyiza-cyama mask igipimo.