Vocoders akwara
Vocoder akwara bụ ihe nlere na-atụgharị ihe nnọchi anya ụda kọmpat, na-abụkarị mel-spectrogram, ka ọ bụrụ ụdị ikuku a na-anụ n'ezie.
Nchịkọta
Ọ bụ ọkwa ikpeazụ nke na-enye ederede ederede na olu nke oge a ụda ha sitere n'okike, nke mmadụ.
Ime miri emi
Ngwakọta okwu ọdịnala na-eji ụda olu na-ahazi mgbama nke na-adakarị ụda mkpọtụ ma ọ bụ rọbọt. Vocoders Neural na-amụta iji wughachi ihe nlele ọdịyo sitere na spectrogram site na ọzụzụ na awa nke ezigbo ndekọ. WaveNet (DeepMind, 2016) bụ ọganihu, na-ebu amụma otu ihe atụ n'otu oge na ihe atụ 16,000+ kwa nkeji, na-emepụta okwu okike na-adọrọ adọrọ ma jiri nwayọọ nwayọọ. Modelsdị ndị ọzọ na-ere ahịa obere akpa ahụ na-agba ọsọ maka ọsọ: WaveGlow jiri ọgbọ na-asọ asọ, Parallel WaveGAN na MelGAN jiri netwọọdụ mmegide na-emepe emepe, na HiFi-GAN ghọrọ ọkọlọtọ ama ama site na iwepụta ụda 22kHz dị elu ngwa ngwa karịa ozugbo. Taa, ụda olu bụ ihe fọrọ nke nta ka ọ bụrụ ọkara nke abụọ nke pipeline abụọ, jikọtara ya na ihe ngosi ụda dị ka Tacotron 2 ma ọ bụ FastSpeech nke na-emepụta mel-spectrogram.
Nghọta nka nka
Otu mel-spectrogram na-atụfu ozi usoro ọdịyo ahụ, na-edobe naanị otu esi ekesa ike n'okirikiri ugboro ugboro ka oge na-aga. Ọrụ siri ike nke ụda olu na-emepụta ụdị ebili mmiri na-ekwekọghị ekwekọ, nke oke ụdịdị dị iche iche dabara na ntinye ahụ. Vocoders dabere na GAN dị ka HiFi-GAN na-eji ọtụtụ ịkpa ókè na-enyocha mgbaàmà n'ọtụtụ dị iche iche na oge dị iche iche, na-amanye onye na-emepụta ọkụ ka ọ mepụta nkọwa dị mma dị ka ihe kwekọrọ na ogologo oge nke consonants.
Mmetụta atụmatụ
Nweta na iru
Ọ na-eme ka nnweta ya dịkwuo mma site na ndegharị, ịkọ akụkọ, na ntụgharị olu.
Ọnụ ego na mmefu ego
Ndị otu mgbasa ozi nwere ike ibubata ọdịyo a na-egbu maramara ngwa ngwa site na iji obere mmefu ego.
Ọsọ na ọnụ ọgụgụ
Sistemụ na-eche ihu ndị ahịa nwere ike hazie mkparịta ụka n'ọtụtụ buru ibu.
Ọdịnihu nke Neural Vocoders
Vocoders na-adị ntakịrị na ngwa ngwa ka ha nwee ike ịgba ọsọ na ekwentị na ngwaọrụ agbakwunyere na-enweghị njikọ igwe ojii. Enwekwara ike na-aga n'ihu vocoders eluigwe na ala nke na-eme ka ọ bụrụ onye na-ekwu okwu, asụsụ, abụ, ma ọ bụ ọbụna ụda na-abụghị nke okwu na-enweghị ọzụzụ. Usoro yiri nke ahụ na-akụda ụda olu ozugbo n'ime sistemu ngwụcha ruo ọgwụgwụ yana codec neural, na-eme ka ahịrị dị n'etiti usoro egwu dị iche iche na nke na-efegharị efegharị na ibelata ihe ndị ewepụtara site na ịgafe spectrogram etiti.
Mmejuputa n'ezie n'ụwa
Ịmepụta ọdịyo ikpeazụ a na-ekwu na ndị enyemaka ederede-gaa-na-okwu dị ka ndị na-agụ ihuenyo na ngwa igodo
Na-emepụta olu emechitere na-ada ụda nke okike n'ime ntinye akwụkwọ na ngwa nkọwa akwụkwọ ọdịyo
Na-ewughachi olu abụ na egwu AI na ngwa-ngwa-ngwa-ngwa
Ịkwado mmepụta olu na ngwaọrụ maka ndị na-ekwu okwu smart yana ngwaọrụ nnweta na-enweghị njegharị gburugburu sava
Ihe ize ndụ & okporo ụzọ nche
Iji olu eme ihe na ihe egwu mpụta ga-abawanye mgbe nkwenye na-efu.
Izi ezi nwere ike ịdaba n'ofe ụda olu, olumba ma ọ bụ gburugburu mkpọtụ.
Enwere ike imehie ọdịyo sịntetik dị ka ezigbo okwu na-enweghị akara doro anya.
Map mmejuputa
Nweta nkwenye doro anya maka ijide olu, imechi, na ijigharị.
Nwale ogo n'ofe ndị na-ekwu okwu dị iche iche yana ọnọdụ ndabere.
Kọwaa mgbe mmadụ ga-enyocha ma ọ bụ kwado nsonye.
Deba aha ọdịyo sịntetik ma debe ndekọ ihe ndekọ maka ịza ajụjụ.
Nọgide na-eme nchọpụta
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 Neural Vocoders 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
Ntuziaka na-esote
HiFi-GAN na GAN Vocoders
Ajụjụ a na-ajụkarị
Kedu ihe bụ Neural Vocoders?
Vocoder akwara bụ ihe nlere na-atụgharị ihe nnọchi anya ụda kọmpat, na-abụkarị mel-spectrogram, ka ọ bụrụ ụdị ikuku a na-anụ n'ezie. Ọ bụ ọkwa ikpeazụ na-enye ederede-ga-okwu na olu nke ọgbara ọhụrụ na-emechi ụda okike ha.
Kedu ọrụ bụ isi nke ụda olu akwara?
Vocoder akwara na-ewere njiri ụda kọmpat, nke na-abụkarị mel-spectrogram, ma na-emepụta ezigbo ụda ụda olu ị na-anụ.
Kedu ụdị 2016 bụ ọganihu nke mere ka ụda olu ụda olu dị ka eke?
WaveNet, sitere na DeepMind na 2016, weputara onyonyo onyonyo site na nlele wee mepụta okwu okike na-adọrọ adọrọ, na-amalite oge vocoder akwara.
Kedu ihe kpatara WaveNet izizi ji nwayọ iwepụta ọdịyo?
WaveNet bụ autoregressive: ihe nlele ọdịyo ọ bụla na-adabere na ndị gara aga, yabụ iwepụta iri puku kwuru puku ihe nlele kwa nkeji bụ ọnụ ọnụ.
Kedu ozi mel-spectrogram na-atụfu nke ọma, na-eme ka ọrụ ụda olu sie ike karị?
Mel-spectrograms na-edobe ịdị ukwuu (ike kwa otu ugboro) mana na-agbada oge, yabụ onye ụda olu ga-ewughachi usoro ebili mmiri na-agbanwe agbanwe nke usoro ya nwere ezi uche.
Kedu ka ndị vocoders dabere na GAN dị ka HiFi-GAN si eme ka eziokwu dị mma?
HiFi-GAN na-eji ọtụtụ ịkpa ókè na-enyocha oge dị iche iche na oge dị iche iche, na-akpali onye na-emepụta ọkụ ka ọ mepụta nkwekọrịta na transients n'ezie.