Meji-Path RNN Iyapa
Meji-Path RNN (DPRNN) jẹ faaji ipinya ohun ohun ti o pin ọna gigun pupọ ti awọn ẹya ohun sinu awọn ṣoki agbekọja kukuru ati ṣe ilana wọn ni awọn ọna yiyan meji nitorina awọn nẹtiwọọki loorekoore le ṣe apẹẹrẹ awọn alaye agbegbe mejeeji ati eto agbaye.
Akopọ
It matters because it made high-quality separation of long recordings practical.
Jin Dive
Awọn nẹtiwọọki loorekoore Ijakadi pẹlu awọn ilana gigun pupọ, ati ohun afetigbọ akoko-akoko ni awọn oṣuwọn iṣapẹẹrẹ giga n ṣe awọn ọna ṣiṣe pẹlu ẹgbẹẹgbẹrun awọn igbesẹ. DPRNN (2020, Luo, Chen, Yoshioka) yanju eyi nipa titunṣe lẹsẹsẹ ẹya sinu akoj 2D ti awọn ege agbekọja. Lẹhinna o paarọ awọn iwe-iwọle RNN meji: awọn awoṣe intra-chunk RNN fun igba kukuru, awọn ilana agbegbe laarin chunk kọọkan, ati awọn awoṣe RNN inter-chunk kan awọn awoṣe awọn igbẹkẹle igba pipẹ kọja awọn chunks. Iṣakojọpọ pupọ ti awọn bulọọki-ọna meji wọnyi jẹ ki awoṣe mu ipo ọrọ ti o gbooro ni gbogbo ọrọ nigba ti RNN kọọkan nikan rii nigbagbogbo ti o le ṣakoso, ferese-ipari-ipari. Ti lọ silẹ sinu ilana Conv-TasNet gẹgẹbi rirọpo fun oluyatọ TCN, DPRNN ṣe jiṣẹ awọn anfani nla ni didara ipinya pẹlu kika paramita iwapọ.
Imọ-imọ-ẹrọ
Ilana bọtini ni ipin pẹlu yiyan ti nwaye. Ọkọọkan gigun ti ipari L ti ṣe pọ sinu matrix ti K chunks ti ipari S (pẹlu 50% ni lqkan). Intra-chunk RNN nṣiṣẹ lẹgbẹẹ S (agbegbe), lẹhinna inter-chunk RNN nṣiṣẹ lẹgbẹẹ K (agbaye), ọkọọkan ni igbagbogbo bidirectional. Nitoripe gbogbo awọn ilana RNN nikan ni awọn igbesẹ S tabi K, iṣapeye duro ni iduroṣinṣin ati aaye gbigba imunadoko di ọkọọkan ni kikun lẹhin awọn bulọọki diẹ. Ni lqkan-afikun reconstructs awọn ọkọọkan.
Ipa Ilana
Wiwọle ati arọwọto
O ṣe ilọsiwaju iraye si nipasẹ transcription, alaye, ati awọn atọkun ohun.
Iye owo ati isuna
Awọn ẹgbẹ Media le firanṣẹ ohun didan yiyara pẹlu awọn isuna-owo kekere.
Iyara ati iwọn
Awọn ọna ṣiṣe ti nkọju si alabara le ṣe ilana awọn ibaraẹnisọrọ sisọ ni iwọn nla.
Ojo iwaju ti Meji-Path RNN Iyapa
Ero-ọna meji-meji DPRNN di awoṣe ti o kọja awọn sẹẹli RNN pato rẹ. SepFormer aṣeyọri nla ti paarọ awọn RNNs fun Awọn Ayirapada inu inu intra/inter chunk be kanna, ati TF-GridNet gbooro si ṣiṣe ọna-meji ni gbogbo akoko ati igbohunsafẹfẹ. Reti ipin-ati apẹẹrẹ miiran lati jẹ idinamọ bulọọki ile boṣewa fun awoṣe ohun afetigbọ-tẹle gigun, pọ si pọ pẹlu akiyesi ati lo kọja ọrọ sisọ si orin ati ipinya ohun gbogbogbo.
Real-World imuse
Iyapa ọpọ awọn agbohunsoke nigbakanna ni ipade pipẹ tabi awọn gbigbasilẹ ifọrọwanilẹnuwo.
Agbara intra/inter-chunk ẹhin nigbamii ti a ṣe atunṣe nipasẹ SepFormer fun iyapa-ti-ti-aworan.
Yiyasọtọ ohun ibi-afẹde kan fun kikọ silẹ ni isalẹ ni ariwo, awọn ibaraẹnisọrọ agbekọja.
Ninu ohun afetigbọ gigun gẹgẹbi awọn ikowe tabi awọn ijiroro nronu nibiti awọn agbọrọsọ sọrọ lori ara wọn.
Awọn ewu & Awọn ọna iṣọ
ilokulo ohun ati awọn ewu afarawe ṣe pọ si nigbati igbanilaaye ba sonu.
Yiye le ju silẹ kọja awọn asẹnti, awọn ede-ede, tabi awọn agbegbe alariwo.
Ohun afetigbọ sintetiki le jẹ aṣiṣe fun ọrọ ododo laisi isamisi to yege.
Ilana Ilana imuse
Gba ifọkansi ti o fojuhan fun gbigba ohun, ti ẹda, ati ilotunlo.
Didara idanwo kọja awọn agbohunsoke oniruuru ati awọn ipo abẹlẹ.
Ṣetumo nigbati eniyan gbọdọ ṣe atunyẹwo tabi fọwọsi awọn abajade.
Aami ohun sintetiki ki o tọju awọn igbasilẹ provenance fun iṣiro.
Tesiwaju Ṣiṣawari
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 Dual-Path RNN Separation 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
Itọsọna atẹle
Awọn awoṣe Onitumọ RNN
Awọn ibeere ti a beere nigbagbogbo
What is Dual-Path RNN Separation?
Meji-Path RNN (DPRNN) jẹ faaji ipinya ohun ohun ti o pin ọna gigun pupọ ti awọn ẹya ohun sinu awọn ṣoki agbekọja kukuru ati ṣe ilana wọn ni awọn ọna yiyan meji nitorina awọn nẹtiwọọki loorekoore le ṣe apẹẹrẹ awọn alaye agbegbe mejeeji ati eto agbaye. O ṣe pataki nitori pe o ṣe iyapa didara giga ti awọn igbasilẹ gigun ti o wulo.
Isoro wo ni Meji-Path RNN yanju ni akọkọ?
DPRNN ṣe atunto awọn ilana-igba pipẹ pupọ si awọn ṣoki ki awọn RNN le mu agbegbe mejeeji ati agbegbe agbaye ṣiṣẹ laisi gige gigun.
Kini awọn ọna meji ni RNN-Path meji?
Awọn ilana RNN kan laarin chunk kọọkan fun awọn ilana agbegbe; awọn ilana miiran kọja awọn chunks fun ọna gigun-gun.
Bawo ni a ṣe pese lẹsẹsẹ ẹya gigun ṣaaju awọn bulọọki ọna meji?
DPRNN ṣe agbo ọkọọkan gigun sinu matrix ti awọn ṣoki agbekọja nitorina RNN kọọkan nikan ṣe ilana window kukuru kan.
Ninu ilana wo ti o wa tẹlẹ ni a ti fi DPRNN silẹ bi aropo oluyapa?
DPRNN rọpo oluyapa TCN inu opo gigun ti epo Conv-TasNet, titọju koodu koodu ti ẹkọ ati oluyipada.
Awoṣe nigbamii wo ni o tọju ọna ọna-meji DPRNN ṣugbọn rọpo awọn RNN pẹlu Awọn Ayirapada?
SepFormer tun lo apẹrẹ intra/inter-chunk meji-ọna apẹrẹ ṣugbọn paarọ awọn sẹẹli loorekoore fun akiyesi ara ẹni Transformer.