UGoogle ukhuphe iV2 yeLyra, ikhowudi yemvelaphi esezantsi yebitrate

ULyra ikhowudi yomsindo kaGoogle

UGoogle ukhuphe inguqulelo yesibini yeLyra, ekumgangatho ophezulu, ephantsi-bitrate codec eyenza unxibelelwano lwelizwi lufumaneke nakwiinethiwekhi ezicothayo.

Mva nje UGoogle utyhilwe ngeposti yebhlog, ikhupha inguqulelo yesibini yecodec yakho yomsindo "Lyra-V2", esebenzisa ubuchule bokufunda ngoomatshini ukufikelela elona lizwi liphakamileyo xa usebenzisa amajelo onxibelelwano acothayo.

Inguqulelo entsha yazisa inguqu kuyilo olutsha lweneural network, inkxaso yamaqonga ongezelelweyo, ulawulo lwebitrate oluphuculweyo, ukuphuculwa komsebenzi, kunye nomgangatho ophezulu wesandi.

Ngoku sikhulula i-Lyra V2, ngoyilo olutsha olonwabela inkxaso yeqonga elibanzi, ibonelela ngesakhono esine-scalable bitrate, ukusebenza ngcono, kunye nomgangatho ophezulu womsindo. Ngolu khululo, sijonge phambili ekuqhubekeni nokuvela kunye noluntu kwaye, ngokuyila kwakho ngokudibeneyo, ubone usetyenziso olutsha oluphuhliswayo kunye nemikhombandlela emitsha evelayo.

Malunga noLyra

Ngokumalunga nomgangatho wedatha yelizwi ehanjiswa ngesantya esiphantsi, I-Lyra iphezulu kakhulu kuneekhowudi zemveli abasebenzisa iindlela zokwenziwa komqondiso wedijithali. Ukuze kuphunyezwe ukuhanjiswa kwelizwi elikumgangatho ophezulu phantsi kweemeko zolwazi oluncinci olugqithisiweyo, ukongeza kuxinzelelo oluqhelekileyo lomsindo kunye neendlela zokuguqula umqondiso, ULyra usebenzisa imodeli yelizwi esekelwe kwinkqubo yokufunda ngomatshini ekuvumela ukuba udale ulwazi olungekhoyo. ngokusekelwe kwiimpawu zentetho eziqhelekileyo.

I-codec ibandakanya i-encoder kunye ne-decoder. I-algorithm ye-encoder Ikhupha iiparamitha zedatha yelizwi rhoqo nge-20 millisecond, izicinezele kwaye idlulisele kumamkeli. phezu kothungelwano ngesantya bit 3,2 kbps ukuba 9,2 kbps.

Kwicala lomamkeli, idikhowuda isebenzisa imodeli yokuvelisa ukwenza kwakhona isiginali yentetho eyintsusa esekwe kwiiparamitha zomsindo ezithunyelwayo, kuqukwa neespectrogram zetshokhwe yelogarithmic ezithathela ingqalelo iimpawu zamandla entetho kudederhu lwamaza ohlukeneyo. .

Yintoni entsha kwiLyra V2?

ULyra V2 usebenzisa imodeli entsha yokuvelisa esekwe kwinethiwekhi ye-SoundStream neural, eneemfuno eziphantsi zokubala, ezivumela ukuchazwa kwe-real-time decoding nakwiinkqubo zamandla aphantsi.

Imodeli esetyenziselwa ukwenza isandi iqeqeshelwe ukusebenzisa amawaka aliqela eeyure zokurekhodwa kwelizwi kwiilwimi ezingaphezu kwama-90 (I-TensorFlow Lite isetyenziselwa ukuqhuba imodeli). Ukusebenza kokuphunyezwa okucetywayo kwanele ukubethelela kunye nokucacisa ilizwi kwii-smartphones zoluhlu lwamaxabiso aphantsi.

Ukongeza ekusebenziseni imodeli eyahlukileyo yokuvelisa, uguqulelo olutsha lukwagqamile ekufakweni kwamakhonkco kunye ne-RVQ quantifier (I-Residual Vector Quantizer) kwi-architecture ye-codec, eyenziwa kwicala lomthumeli ngaphambi kokuhanjiswa kwedatha, kunye necala lomamkeli emva kokufumana idatha.

I-quantizer iguqula iiparameters ezibonelelwe yi-codec kwiiseti zeepakethi, i-encoding ulwazi malunga nomlinganiselo okhethiweyo we-bit. Ukuqinisekisa amanqanaba omgangatho owahlukileyo, i-quantizers ibonelelwa ngee-bitrate ezintathu (3,2kbps, 6kbps, kunye ne-9,2kbps), kokukhona i-bitrate iphezulu, kokukhona umgangatho ungcono, kodwa kokukhona uphezulu iimfuno ze-bandwidth.

uyilo olutsha iye yanciphisa ukulibaziseka kokuhanjiswa kwesignali ukusuka kwi-100 millisecond ukuya kwi-20 milliseconds. Ukuthelekisa, i-Opus codec yeWebRTC ibonise ukulibaziseka kwe-26,5 ms, 46,5 ms, kunye ne-66,5 ms kumazinga e-bit avavanyiwe. Ukusebenza kwe-encoder kunye ne-decoder nako kunyuke kakhulu: Xa kuthelekiswa nenguqulo yangaphambili, kukho ukukhawuleza ukuya kumaxesha angama-5. Ngokomzekelo, kwi-smartphone ye-Pixel 6 Pro, i-codec entsha i-codec kunye ne-decodes isampuli ye-20ms kwi-0,57ms, ephindwe ngama-35 ngokukhawuleza kunokuba ifuneka kwi-real-time streaming.

Ukongeza kwintsebenzo, siye sakwazi ukuphucula umgangatho wokubuyisela isandi: ngokomlinganiselo we-MUSHRA, umgangatho wentetho kwi-bit rates ye-3,2 kbps, 6 kbps kunye ne-9,2 kbps xa usebenzisa i-codec ye-Lyra V2 ihambelana nemilinganiselo ye-bit ye-10 kbps, 13 kbps kunye ne-14 kbps xa usebenzisa i-Opus codec.

Gqibela ukuba unomdla wokwazi okungakumbi ngayo, ungazijonga iinkcukacha kwi eli khonkco lilandelayo.


Shiya uluvo lwakho

Idilesi yakho ye email aziyi kupapashwa. ezidingekayo ziphawulwe *

*

*

  1. Uxanduva lwedatha: UMiguel Ángel Gatón
  2. Injongo yedatha: Ulawulo lwe-SPAM, ulawulo lwezimvo.
  3. Umthetho: Imvume yakho
  4. Unxibelelwano lwedatha: Idatha ayizukuhanjiswa kubantu besithathu ngaphandle koxanduva lomthetho.
  5. Ukugcinwa kweenkcukacha
  6. Amalungelo: Ngalo naliphi na ixesha unganciphisa, uphinde uphinde ucime ulwazi lwakho.