Mgbakwunye WaveGAN Vocoder
Parallel WaveGAN bụ vocoder akwara ngwa ngwa nke na-atụgharị mel-spectrogram ka ọ bụrụ usoro ebili mmiri dị egwu site na iji obere GAN, na-ewepụta nlele niile ozugbo.
Nchịkọta
It matters because it gives near-real-time, high-quality speech with a compact model.
Ime miri emi
Vocoder bụ ọkwa ikpeazụ nke pipeline TTS: ọ na-atụgharị maapụ njirimara ụda (na-abụkarị mel-spectrogram) n'ime ụda ụda ị na-anụ. Parallel WaveGAN, nke Yamamoto, Song na Kim tụpụtara na 2019, na-eme nke a site na onye na-emepụta ụdị WaveNet na-abụghị nke a zụrụ azụ dị ka netwọk mmegide. Kama ịkọ otu ihe atụ ọdịyo n'otu oge dị ka WaveNet mbụ, ọ na-emepụta ụdị ebili mmiri niile n'otu oge, na-eme ka ọ dị ngwa ngwa. Ntụziaka ya bụ isi na-ejikọta ọnwụ adversarial yana mfu nke obere oge Fourier transform (STFT), yabụ ihe nlereanya ahụ dabara n'ezie mgbama gafee ọtụtụ oge na akpịrịkpa ugboro. Nsonaazụ bụ obere jenerato (ihe dị ka nde 1.4) nke na-agba ọsọ ọsọ ọtụtụ ugboro karịa ozugbo na GPU.
Nghọta nka nka
Igwe ọkụ bụ netwọk gbasasịrị-convolution nke etinyere na mel-spectrogram yana ntinye mkpọtụ, mkpọtụ mkpọtụ gbakwunyere atụmatụ ozugbo na nlele. Ọzụzụ jikọtara ọnụ na-ebelata mfu STFT ọtụtụ mkpebi, gbakọọ site n'ịtụle nnukwu spectrogram n'ọtụtụ FFT na ogologo hop, yana mfu nke onye na-akpa ókè na-ekpebi eziokwu. Okwu STFT na-eme ka ọ kwụsie ike ma na-eme ka ọzụzụ ọgụ dị ngwa, na-ewepụta ma nkọwa zuru oke yana ụdị ọdịdị sara mbara na-enweghị mgbagha.
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 Parallel WaveGAN Vocoder
Parallel WaveGAN nyere aka guzobe vocoders GAN dị ka ndabara bara uru, yana mfu STFT nke ọtụtụ mkpebi ya na-apụta ugbu a n'ofe ndị ga-anọchi anya HiFi-GAN na ọtụtụ sistemụ nkwanye. Ntugharị a na-atụ aka n'ebe ụda olu dị obere, nke dị ala maka ndị na-enyere ngwaọrụ aka, ihe enyemaka ịnụ ihe, na ntụgharị olu dị ndụ, gbakwunyere vocoders zuru ụwa ọnụ nke na-eme ka ndị na-ekwu okwu na-adịghị ahụ anya. Na-atụ anya mwekota siri ike na TTS njedebe na njedebe na ntinye nke ọma na mkpanaka na ibe agbakwunyere.
Mmejuputa n'ezie n'ụwa
Mbupute okwu n'ezie na ndị enyemaka olu mkpanaka ebe nkwụsịtụ na nha ụdị dị mkpa
Na-eje ozi dị ka onye na-ebupụta waveform jikọtara ya na ụdị egwu dị ka Tacotron 2 ma ọ bụ FastSpeech
ederede-gaa-okwu ngwaọrụ maka ngwaọrụ nnweta na-enweghị ike ịdabere na igwe ojii
Sistemụ ntụgharị olu na-emegharị spectrograms tụgharịrị ka ọ bụrụ ọdịyo na-ada ụda eke
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 Parallel WaveGAN Vocoder 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
MelGAN Generative Vocoder
Ajụjụ a na-ajụkarị
What is Parallel WaveGAN Vocoder?
Parallel WaveGAN bụ vocoder akwara ngwa ngwa nke na-atụgharị mel-spectrogram ka ọ bụrụ usoro ebili mmiri dị egwu site na iji obere GAN, na-ewepụta nlele niile ozugbo. Ọ dị mkpa n'ihi na ọ na-enye nso nso-ezigbo, okwu dị elu na kọmpat nlereanya.
Gịnị bụ ọrụ nke ụda olu dị ka Parallel WaveGAN?
Vocoder bụ ọkwa ikpeazụ TTS nke na-ejikọta ụda ụda n'ezie site na njirimara ụda dị ka mel-spectrogram.
Kedu ka Parallel WaveGAN si ewepụta nlele ọdịyo ma e jiri ya tụnyere WaveNet mbụ?
Parallel WaveGAN bụ onye na-abụghị autoregressive, na-emepụta dum waveform otu ugboro kama sample site sample, nke na-eme ka ọ dị ngwa ngwa.
Kedu ọnwụ dị n'etiti Parallel WaveGAN ma e wezụga ọnwụ iro?
Ọ na-ejikọta mfu adversarial yana mfu STFT nwere ọtụtụ mkpebi nke na-atụnyere ịdị ukwuu spectrogram n'ọtụtụ akpịrịkpa.
Kedu ụlọ ọrụ Parallel WaveGAN generator na-eji?
Onye na-emepụta ọkụ bụ netwọkụ WaveNet nke arụnyere site na mgbagha gbasasịrị, nwere ọnọdụ na mel-spectrogram na mkpọtụ.
Kedu ihe kpatara Parallel WaveGAN ji mara mma maka ibuga?
N'ihe fọrọ nke nta ka ọ bụrụ nde 1.4, ọ dị obere ma na-agba ọsọ ọtụtụ oge karịa ozugbo, ọ dị mma maka iji ngwaọrụ.