TWS headset isignali yedijithali ADM
Ngokukhula okuqhubekayo kwemakethe yamahedisethi e-TWS (True wireless stereo). Izidingo zabasebenzisi zolwazi lomkhiqizo nazo zithuthukisiwe zisuka kuzixhumanisi ezisheshayo zaya kumazinga aphezulu. Isibonelo, kusukela kulo nyaka, inani elikhulu lamahedisethi e-TWS anamakholi acacile avele emakethe.
Ukuze unike amandla ukuxhumana kwezwi okucacile ezindaweni ezinomsindo kakhulu, kungenzeka yini ukukhiqiza izikimu ezihlanganisa amasignali asuka endlebeni yangaphakathi kanye nemakrofoni yangaphandle ukuze kusetshenziswe ubuchwepheshe bokuhlanganisa ibhande elincane elihlakaniphile, elivumelana nemvelo. Eqinisweni, ezinye izinkampani ze-algorithm zasekhaya nezangaphandle zizinikele kulokhu, futhi zithole imiphumela ethile.
Yiqiniso, izinkampani eziningi zezixazululo manje sezigcizelele ngokukhethekile izixazululo zokunciphisa umsindo wezingcingo ezifana ne-AI enqenqemeni (lena eyodwa), kodwa empeleni, ilungiselelwe kangcono izixazululo ezikhona zokunciphisa umsindo wezingcingo, ngakho-ke le ngxenye isusiwe, ake sibheke ezinye izingxenye eziyisisekelo kuqala Isingeniso, okungukuthi, lokho okungenziwa ukuncishiswa komsindo.
Sekukonke, ukunciphisa umsindo wezingcingo kuncike ekuvumelaneni kwe-Uplink (uplink) ne-Downlink (downlink). Cishe i-Microphone Array/AEC/NS/EQ/AGC/DRC, ubudlelwano obuphusile bumi kanje:
I-ADM (I-Adaptive Directional Microphone Array) iwubuchwepheshe bokucubungula isignali yedijithali eyenza imakrofoni eqondisayo noma ekhansela umsindo kusetshenziswa imibhobho emibili kuphela ye-omnidirectional. I-ADM ishintsha ngokuzenzakalelayo izici zayo zokuqondisa ukuze inikeze ukuncishiswa komsindo okuphelele ezindaweni ezihlukahlukene kuyilapho igcina ikhwalithi yesignali eyanele. Inqubo yokuguquguquka iyashesha, inokukhetha okunamandla kwefrikhwensi, futhi ingaqeda iziphazamiso eziningi ngesikhathi esisodwa.
Ngokungeziwe ezicini zayo ezinhle zokuqondisa, ama-ADM angenwa kalula nomsindo womoya kunemibhobho yokuqondisa ye-acoustic yendabuko. Ubuchwepheshe be-ADM buvumela izinhlobo ezimbili zokucushwa kwemakrofoni: "endfire" kanye "nomlilo omkhulu".
Ekucushweni kwe-endfire, umthombo wesignali (umlomo womsebenzisi) uku-eksisi (umugqa oxhuma imakrofoni emibili). Ekucupheni okubanzi, iqondise umugqa oqondile ku-eksisi evundlile.
Ekucushweni kwe-endfire, i-ADM inezindlela ezimbili zokusebenza; “i-far-talk” kanye “ne-close-talk”. Kumodi ye-Far-pass, i-ADM isebenza njengombhobho oqondile oqondile, inciphisa isignali kusukela ngemuva nasezinhlangothini ngenkathi igcina isignali kusukela ngaphambili. Kumodi yokukhuluma eduze, i-ADM isebenza njengombhobho ongcono kakhulu osusa umsindo, osusa imisindo ekude ngempumelelo. Inkululeko elinganiselwe yokuklama kwe-acoustic yenza ama-ADM alungele omakhalekhukhwini, okuvumela ukushintshashintsha “okuthambile” phakathi kwezipikha ezikude nezipikha eziseduzane. Kodwa-ke, uma lolu hlobo lomklamo lusetshenziswa kuma-earphone, ikakhulukazi ama-earphone e-TWS, kukhawulelwe kakhulu ukuthi umsebenzisi ulugqoka ngendlela efanele yini. Ngokufanayo nama-airpod, umbhali uye waphawula ukuthi abantu abaningi banazo “zonke izinhlobo ezixakile” izindlela zokugqoka emgwaqweni ongaphansi komhlaba, ezinye zazo okuyizindlebe zabasebenzisi. Umumo, kanye nemikhuba ethile yokugqoka, kubangela ukuthi i-algorithm ingasebenzi esimweni esifanele.
I-Acoustic Echo Khansela (AEC)
Uma ingxenye yesignali eku-duplex (indlela eyodwa ngesikhathi esisodwa) ukuxhumana ibuyela kusiginali yomthombo, ibizwa ngokuthi “echo”. Ku-analogi yebanga elide futhi cishe kuzo zonke izinhlelo zokuxhumana zedijithali, ngisho namasiginali amancane e-echo angabangela ukuphazamiseka ngenxa yokubambezeleka okunzima kohambo lokuya nokubuya.
Kutheminali yokuxhumana yezwi, ama-acoustic echoes akhiqizwa ngenxa yokuhlangana kwe-acoustic phakathi kwesipikha nemakrofoni. Ngenxa yokucutshungulwa okungaqondile okusetshenziswe kusiteshi sokuxhumana, njengamavoda alahlekile kanye ne-transcoding, ama-acoustic echoes kufanele acutshungulwe endaweni (akhanselwe) ngaphakathi kwedivayisi.
I-Noise Suppressor (NS)
Ubuchwepheshe bokucindezela umsindo bunciphisa umsindo omile nowesikhashana kumasignali enkulumo yesiteshi esisodwa, buthuthukisa isilinganiso sesignali-kumsindo, buthuthukisa ukuqondakala kwenkulumo, futhi kunciphisa ukukhathala kokuzwa.
Yebo, ziningi izindlela eziqondile kule ngxenye, njenge-BF (Beamforming), noma i-PF (Isihlungi sokuthunyelwe) nezinye izindlela zokulungisa. Ngokuvamile, i-AEC, NS, BF, ne-PF izingxenye eziyinhloko zokunciphisa umsindo wezingcingo. Kuyiqiniso ukuthi umhlinzeki ngamunye wesixazululo se-algorithm unezinzuzo ezihlukile kanye nokubi.
Kuhlelo lokuxhumana lwezwi olujwayelekile, izinga lesignali yezwi lingahluka kakhulu ngenxa yebanga phakathi komsebenzisi nemakrofoni, nangenxa yezimpawu zesiteshi sokuxhumana.
I-Dynamic Range Compression (DRC) iyindlela elula yokulinganisa amazinga esignali. Ukucindezela kunciphisa ububanzi obuguqukayo besignali ngokunciphisa (ukucindezela) izingxenye zenkulumo ezinamandla kuyilapho kulondoloza ngokwanele izingxenye zenkulumo ezibuthakathaka. Ngakho-ke, isignali yonke ingakhuliswa ngokwengeziwe ukuze izimpawu ezibuthakathaka zizwakale kangcono.
Ubuchwepheshe be-AGC bukhuphula ngokwedijithali ukuzuza kwesignali (ukukhulisa) uma isignali yezwi ibuthakathaka, futhi iyiminyanisa lapho isignali yezwi inamandla. Ezindaweni ezinomsindo, abantu bavame ukukhuluma kakhulu, futhi lokhu kusetha ngokuzenzakalelayo ukuzuza kwesiteshi semakrofoni kube inani elincane, ngaleyo ndlela kunciphisa umsindo ozungezile kuyilapho kugcina izwi lentshisekelo lisezingeni elifanele. Futhi, endaweni ethulile, abantu bakhuluma ngokuthula ukuze amazwi abo akhuliswe i-algorithm ngaphandle komsindo omkhulu.
Isikhathi sokuthumela: Jun-07-2022
