Koeffisien cepstral yu am fréquence mel
Koeffisien Cepstral Mel-Fréquence (MFCCs) ay lim yu ndaw lañu yuy tënk jëmmu spectre fréquence bu son ci anam wi nopp nit diko gisee.
Résumé
For decades they were the workhorse feature for speech recognition, speaker identification, and music analysis.
Plongeur bu xóot
MFCC yi dañuy soppi benn wàll audio bu gàtt ci lu tollu ci 13 nimero yuy jàpp timbre bi. Tuyo bi dafay jël forme onde bi, daal di koy xaaj ~25ms, xayma spectre puissance jaaraleko ci transformation Fourier, ba noppi deforme axis fréquence bi ci scale mel, biy xaaj bande yi ci anam wi cochlea di defee: ci suufu 1kHz ak ci kaw. Energie mel yi dañu leen komprime ci log (di toppandoo ni ñuy gisee kàddu yi) ba noppi ñu mujjee jaar ci coppite cosine bu diskret, bi leen di decorreler ba noppi dajale xibaar yi ci koeffisien yu néew yi njëkk. Resultaa bi dafay dëgër ci bruit bi ak tonu waxkat bi, moo tax sistemu wax yu Hidden Markov ak Gaussian Mixture Model dañu daan sukkandikoo ci MFCCs daanaka ci àdduna bi yépp balaa ñuy jàng lu xóot.
Gis-gis xarala
Balaas mel bi mingi méngoo ak ni ñuy gisee ton bi ak mel = 2595 log10(1 + f/700), kon jéego mel yu tolloo dañuy tolloo diggante. Transform cosine diskret bi mujj (DCT) mooy jéego 'cepstral' bi: dafay jëfandikoo spectre log-mel bi ni siñaal ba noppi tàqale forme tract vocal biy soppiku ndànk (coefficients cepstral yu woyof, wàll wi ñuy denc) ak armonic yu gaaw yi (coefficients yu kawe, dañuy faral di sànni), ràññeekaay bu baax ci pitch foneker.
njeextalu pexe
Dugg ak yegg
Dafay gëna yombal jëfandikoo gi jaaraleko ci transkripsioŋ, nettali ak interfaasu baat.
Njëgg ak budget
Ekipu mejaa yi mën nañu yónnee audio bu leer ci anam wu gëna gaaw te seen xaalis gëna néew.
Gaawaay ak yaatuwaay
Sistem yiy jàkkarloo ak kiliyaan bi mën nañu def waxtaan ci anam wu gëna yaatu.
Ëlëgu koeffisien Cepstral Mel-Fréquence
Reseau yu xóot yi dem ba ci njeexte gi dañuy gëna jàng màndarga yi jub ci forme onde yu ñor wala spectrogram log-mel, di sànni DCT, kon MFCC yu sell yi dañuy dem ci ASR bi gëna xarañ. Waaye ñu ngi wéy di siiw ci liggéey yu woyof, ci aparey, ak yu néew done: gis baatu-caabi, gis liggéeyu baat, emprent audio, ak bioakustik. Xaarandi MFCCs ñu wéy di nekk baseline bu baax, buñu mëna tekki doonte jàng front-ends yi ñoo ëpp doole ci model yu mag yi.
Doxal ci àdduna dëgg
Man-mani akustik ngir xàmmeekatu kàddu HMM-GMM yu melni sistem Sphinx ak HTK yu njëkk ya
Saytu kàddukat bi ak bind ci surnaal, ràññee kiy wax ci woote
Xeetu music ak emprent way (mënngoo timbre ci stil Shazam)
Gis njuumte masin wala woote baayima ci audio ci surveillance usine ak bioakustik
Risk yi ak balustrade yi
Jëfandikoo baat ci anam wu jaarul yoon ak niru ak nit dafay gëna yokk sudee nanguwul.
Jaar-jaar mën na wàññeeku ci aksan yi, dialect yi wala barab yu bari xumbaay.
Audio synthetik mën nañu ko jaawale ak wax ju dëggu sudee amul etiket bu leer.
Roadmap ngir samp gi
Wutal ndigal bu leer ngir jàpp baat bi, klone ko ak jëfandikoowaat ko.
Saytu kalite ci kàddukat yu bari ak anam yu bari ci ginaaw.
Mandargal kañ la nit wara xoolaat wala nangu ay génne.
Etiketu audio synthetik te nga denc dokimaa ci fimu bawoo ngir mëna lim.
Weyal di banneexu
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 Mel-Frequency Cepstral Coefficients 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
Gis bi ci topp
Spektrogram Mel
Laaj yi ñuy faral di laaj
What is Mel-Frequency Cepstral Coefficients?
Koeffisien Cepstral Mel-Fréquence (MFCCs) ay lim yu ndaw lañu yuy tënk jëmmu spectre fréquence bu son ci anam wi nopp nit diko gisee. Lu weesu fukki at ci ginaaw lañu nekkoon lu gëna am solo ci xàmmee kàddu yi, xàmmee waxkat yi ak jàngat music bi.
Luy 'mel' bi ci MFCC?
Mel scale bi dafay deforme frequency bi suko defee jéego yu tolloo yi sori ñu tolloo seen biir ci nit ñi, ñu tàqaloo bu baax ci frequency yu woyof yi ak yu ñaw yi ci yu kawe yi.
Ban coppite lañu mujjee jëfandikoo ngir am koeffisien cepstral yi?
Ginaaw ñu xaymaa energie log-mel yi, ab coppite cosinus bu diskret moo leen di decorrelé ba noppi boole ay leeral ci coefficient yu néew yi njëkk.
Lan moo waral MFCC yi gëna am doole ci kàdduy waxkat bi?
Koeffisien cepstral yu woyof yi dañuy jàpp enveloppe tract vocal bi (ëmbiit fonetik bi) waaye yu kawe yi dañuy jàpp ton bi, kon tëye yu woyof yi dañuy dindi ton bi.
Lu tollu ci ñaata koeffisien MFCC lañuy faral di denc ci kaadar bu nekk?
Tanneef bu ñuy gëna faral di tànn mooy lu tollu ci 13 koeffisien, yenn saa mu yokk ci seeni delta, yuy tënk timbre bi bu baax.
Lu tax ñuy jëfandikoo log ci energie mel-band?
Li nit ñi di gis ci kàddu yu kawe yi dafay nuru lu logarithmik, kon kompresioŋ log dafay tax man-man yi gëna méngoo ak gis-gis bi ba noppi dakkal rang dynamique bi.