Gane Kakakin Kakakin ECAPA-TDNN
ECAPA-TDNN shine gine-ginen cibiyar sadarwa na jijiyoyi wanda ke juya kowane shirin magana zuwa ƙarami 'burin murya', yana ba injina damar faɗin wanda ke magana.
Dubawa
It set the state of the art for speaker verification and remains the workhorse behind voice ID systems today.
Zurfafa nutsewa
ECAPA-TDNN yana tsaye ne don Ƙaddamar da Hankalin Tashoshi, Yadawa da Tari a cikin Tsare-tsare na Neural Networks, wanda Desplanques da abokan aiki suka gabatar a cikin 2020. Yana ginawa a kan tsohuwar tsarin x-vector amma yana ƙara haɓaka maɓalli guda uku: Squeeze-Excitation blocks that reweight feature channels, Multi-layer Shallow features, Multi-layer, Shallow Layer and GG. Tashoshi-da-context-dogaran kididdiga mai hankali wanda ke taƙaita magana mai tsayi mai tsayi zuwa madaidaiciyar vector guda ɗaya. An horar da shi tare da asarar-margin softmax (AAM-softmax) akan manyan kamfanoni kamar VoxCeleb, yana samar da abubuwan haɗawa inda shirye-shiryen bidiyo iri ɗaya suka taru sosai. Ana kwatanta hotunan murya biyu da kamancen cosine. A kan gwajin VoxCeleb1 ya saita ƙimar kuskure daidai da kusan kashi 1, babban tsalle akan tsarin da suka gabata.
Fahimtar Fasaha
Babban abin zamba shine haɗar kididdiga mai hankali: maimakon kawai matsakaicin fasalulluka na matakin firam, hanyar sadarwar tana koyon ma'aunin hankali kowane tashoshi don haka firam ɗin firam (maganganun bayyanannen magana) suna ƙidayawa fiye da shiru ko amo, sannan ta ƙididdige duka ma'ana mai nauyi da daidaitaccen ma'auni. Abubuwan toshe SE da Res2Net-style juzu'ai masu yawa suna ba da damar kowane yanayin yanayin yanayin furci na duniya. Ƙarshe na ƙarshe shine yawanci girma 192, wanda aka zira ta tazarar cosine.
Dabarun Tasiri
Shiga ku isa
Yana inganta samun dama ta hanyar rubutu, ba da labari, da mu'amalar murya.
Kudin da kasafin kuɗi
Ƙungiyoyin kafofin watsa labaru na iya jigilar sauti mai gogewa cikin sauri tare da ƙaramin kasafin kuɗi.
Gudu da sikelin
Tsarin fuskantar abokin ciniki na iya aiwatar da hulɗar magana a mafi girman ma'auni.
Makomar Gane Kakakin Kakakin ECAPA-TDNN
Bincike yana tafiya zuwa gaba-gaba mai kulawa da kai kamar WavLM da wav2vec 2.0 ciyar da ECAPA-salon baya-ƙarshen, wanda ke yanke bayanan da ake buƙata da kuma haɓaka ƙarfi ga amo da gajerun shirye-shiryen bidiyo. Yi tsammanin haɗin kai mai ƙarfi tare da hana zubewa don haka ƙirar guda ɗaya ta gano kuma ta tabbatar da mai magana, ƙananan juzu'i don amfani da na'urar, da ingantaccen aiki mai ƙarfi don rage gibin kurakurai a cikin lafazin lafazin, shekaru, da harsuna yayin da ƙirar ƙirar murya ta faɗaɗa cikin banki da ikon samun dama.
Aiwatar da Gaskiyar Duniya
Shigar biometric na murya don banki na tarho, inda hoton muryar mai kiran ya dace da samfuri mai rijista maimakon PIN.
Ƙaddamar da lasifikar magana a cikin haɗuwa da kayan aikin kwafi, yiwa lakabin 'wanda yayi magana lokacin' ta hanyar tara abubuwan ECAPA.
Tabbatar da lasifika na gaban shari'a da cibiyar kira don tuta ko rikodin biyu sun fito daga mutum ɗaya.
Ƙaddamar da girke-girke-tabbacin magana a cikin buɗaɗɗen kayan aiki kamar SpeechBrain da Kaldi don masu bincike da farawa.
Hatsari & Tsare-tsare
Rashin amfani da murya da haɗarin kwaikwaya yana ƙaruwa lokacin da aka rasa izini.
Daidaituwa na iya faɗuwa cikin lafuzza, yaruka, ko mahalli masu hayaniya.
Ana iya kuskuren sauti na roba don ingantacciyar magana ba tare da bayyananniyar lakabi ba.
Taswirar Hanya
Sami tabbataccen izini don ɗaukar murya, cloning, da sake amfani.
Gwajin ingantattun masu magana daban-daban da yanayin baya.
Ƙayyade lokacin da dole ne ɗan adam ya duba ko ya amince da abubuwan da aka fitar.
Yi lakabin sauti na roba da kuma adana bayanan da aka tabbatar don yin lissafi.
Ci gaba da Bincike
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 ECAPA-TDNN Speaker Recognition quiz
Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
Jagora na gaba
Gane Audio Chord
Tambayoyin da ake yawan yi
What is ECAPA-TDNN Speaker Recognition?
ECAPA-TDNN shine gine-ginen cibiyar sadarwa na jijiyoyi wanda ke juya kowane shirin magana zuwa ƙarami 'burin murya', yana ba injina damar faɗin wanda ke magana. Ya saita yanayin fasaha don tabbatar da lasifika kuma ya kasance dokin aiki a bayan tsarin ID na murya a yau.
Menene farkon fitowar samfurin ECAPA-TDNN don shirin magana?
ECAPA-TDNN tana yin taswirorin magana mai tsayi mai canzawa zuwa cikin tsayayyen tsayi mai tsayi wanda ke wakiltar halayen muryar mai magana.
Yaya nau'ikan ECAPA-TDNN guda biyu yawanci idan aka kwatanta da yanke shawara idan sun kasance na mai magana ɗaya?
Bayan an fitar da abubuwan sakawa, kamannin cosine (ko maƙiya mai alaƙa kamar PLDA) tana auna yadda muryoyin biyu ke kusa.
Menene Layer na 'kididdiga mai hankali' ke yi?
Haɗin ƙididdiga mai hankali yana ba da ma'aunin hankali da aka koya ga firam ɗin kuma ya haɗa su zuwa ma'auni mai nauyi da daidaitaccen karkata, yana mai da hankali ga firam ɗin bayanai.
Wanne saitin bayanai ne aka fi amfani da shi don horarwa da ƙirar ƙirar magana ta ECAPA-TDNN?
VoxCeleb, babban saitin shirye-shiryen hirar mashahuran da aka goge daga bidiyo, shine ma'auni na horo da kimanta tsarin tabbatar da lasifika.
Menene toshe 'SE' (Squeeze-Excitation) ke ba da gudummawa ga gine-gine?
Matsi-Excitation tubalan koyi ma'auni kowane tasha alama ta amfani da bayanan duniya, barin cibiyar sadarwa jaddada mafi amfani tashoshi.