Eji RNnoise jụ ikwu okwu
RNnoise bụ obere netwọkụ akwara ozi dị ngwa nke na-ewepụ mkpọtụ ndabere site na okwu ozugbo.
Nchịkọta
Created by Xiph.Org's Jean-Marc Valin, it pairs classic signal processing with a small recurrent network so it runs on ordinary CPUs and even embedded devices.
Ime miri emi
RNnoise, ewepụtara na 2017, ka emebere maka mbelata mkpọtụ mkpọtụ na oku olu. Kama ịmụta ihe niile site na njedebe ruo na njedebe, ọ na-ekewa okwu n'ime ihe dị ka 22 ugboro ugboro nke e mere na ntị mmadụ (ọnụọgụ dị ka Bark) ma na-eji netwọk neural na-emegharị ugboro ugboro na Gated Recurrent Units iji tụọ uru (0 ruo 1) maka band ọ bụla n'otu etiti. Uru ndị ahụ na-ebelata ụgbụ ndị na-eme mkpọtụ ka ha na-edobe ìgwè ndị na-ekwu okwu na-emebibeghị. Ihe nzacha pitch agbakwunyere na-ehichapụ mkpọtụ fọdụrụnụ n'etiti nkwekọ nke okwu a na-ekwu. Ihe nlereanya ahụ niile nwere ihe dị ka nha 85,000, na-agba ọsọ ọsọ karịa ozugbo na otu isi CPU, ma bụrụ ebe mepere emepe n'okpuru ikikere BSD, ya mere etinyere ya n'ọrụ dịka Opus codec ecosystem, Mumble, na OBS Studio.
Nghọta nka nka
Nhọrọ nke isi ihe na-arụ ọrụ na uru band nke nghọta kama ịbụ nnukwu bins spectral raw. Site n'ịkọ amụma naanị ~22 uru n'otu etiti, netwọk GRU na-adị ntakịrị ma na-ezere ihe egwu egwu na-emekarị na ụzọ mwepu ihe nkiri ochie. Njirimara ejiri aka mee (ike band, oge pitch, njikọ pitch) na-enye netwọk nri, na-ejikọta ihe ọmụma DSP na mmụta. Mwepụta olu dị iche iche na-enyere aka nweta uru ọnụ ụzọ ámá n'oge okpokolo agba dị ọcha.
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 ikwu okwu na RNnoise
RNnoise kpaliri ebili mmiri nke ọrụ nkwalite ozugbo; nyocha onye ga-anọchi ya (PercepNet, DeepFilterNet) na-akwalite ịdịmma ya ka ọ na-edobe mmefu ego CPU obere. Na-atụ anya ka ndị denoisers tinye ozugbo n'ime ekweisi, ihe enyemaka ịnụ ihe, na ibe ọgbakọ, ijikọ ya na kagbuo nkwughachi na nkwuwa okwu, na iji echiche nghọta na ọbụna ebumnobi mmepụta. Ngwakọta DSP-gbakwunyere-obere-netwọk uzommeputa na-anọgide na-enwe mmetụta n'ebe ọ bụla obere latency, obere ike, na oghe-isi iyi ihe ikike karịa raw nlereanya size.
Mmejuputa n'ezie n'ụwa
Na-egbochi mkpọsa ahụigodo na ịma jijiji n'oge oku vidiyo na ngwa na-ejikọta RNnoise.
Na-ehicha igwe okwu iyi na OBS Studio site na nzacha mkpọtụ mkpọtụ RNnoise arụnyere n'ime ya.
Imelite nghọta nke nkata olu na egwuregwu yana ngwaọrụ VoIP dị ka Mumble na ngwaike dị ala.
Na-ahazi ndekọ mpaghara mkpọtụ ka njirimara okwu dị n'okpuru na-enweta mgbama dị ọcha karị.
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 Speech Denoising with RNNoise 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
Nkewa okwu na nsogbu Party Cocktail
Ajụjụ a na-ajụkarị
What is Speech Denoising with RNNoise?
RNnoise bụ obere netwọkụ akwara ozi dị ngwa nke na-ewepụ mkpọtụ ndabere site na okwu ozugbo. Mepụtara Xiph.Org's Jean-Marc Valin, ọ na-ejikọta nhazi mgbaàmà kpochapụ na obere netwọkụ na-emegharị ugboro ugboro ka ọ na-agba ọsọ na CPU ndị nkịtị na ọbụna ngwaọrụ agbakwunyere.
Kedu ihe RNnoise na-ebu amụma maka ụda olu ọ bụla dị mkpụmkpụ?
RNnoise na-atụle uru otu-band na-eme ka ụda mkpọtụ na-achịkwa ma na-echekwa ndị na-ekwu okwu, kama ịrụ ọrụ n'ụdị ọ bụla.
Kedu ụdị netwọkụ akwara ozi dị na isi RNnoise?
RNnoise na-eji netwọk akwara na-emegharị ugboro ugboro nke ejiri Gated Recurrent Units rụọ, na-enye ya ebe nchekwa nwa oge ka ọ na-adị obere.
Kedu ihe kpatara RNnoise ji eji eriri elekere nghọta kama iji bns spectral raw?
Ịkọ amụma naanị ~ 22 uru band na-eme ka netwọk ahụ dị ntakịrị, ngwa ngwa na nke na-adịchaghị mfe maka ihe egwu egwu na-emebi ụzọ ọ bụla.
Kedu ka ụdị RNnoise siri buru ibu?
RNnoise nwere n'usoro nke 85,000 dị arọ, pere mpe iji na-agba ọsọ ọsọ karịa ozugbo na otu isi CPU.
Kedu ọrụ nke nzacha pitch na RNNoise?
Ihe nzacha comb/pitch na-eji usoro nkwekọ nke okwu ekwupụtara eme ihe iji gbochie mkpọtụ ịnọdụ n'etiti ihe nkwekọ, na-emeju uru ndị otu.