G - Physics – 10 – L
Patent
G - Physics
10
L
G10L 17/00 (2006.01)
Patent
CA 2184256
The present invention relates to a speaker recognition method and system which applies adaptive component weighting to each frame of speech for attenuating non-vocal tract components and normalizing speech components. A linear predictive all pole model is used to form a new transfer function having a moving average component. A normalized spectrum is determined from the new transfer function. The normalized spectrum is defined having improved characteristics for speech components. From the improved speech components, improved speaker recognition over a channel is obtained.
La présente invention concerne une technique et un système d'identification de la personne qui parle, faisant intervenir une pondération adaptive des composantes pour chaque trame de parole afin d'atténuer les composantes étrangères à l'appareil vocal et de normaliser les composantes vocales. On utilise un modèle omnipolaire de prédiction linéaire pour former une nouvelle fonction de transfert ayant une composante moyenne mobile. Le spectre normalisé présente des caractéristiques améliorées au niveau des composantes vocales. Ces composantes vocales améliorées facilitent l'identification de la personne qui parle sur un canal.
Assaleh Khaled T.
Mammone Richard J.
Norton Rose Or S.e.n.c.r.l. S.r.l./llp
Rutgers The State University Of New Jersey
LandOfFree
Speaker identification and verification system does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Speaker identification and verification system, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Speaker identification and verification system will most certainly appreciate the feedback.
Profile ID: LFCA-PAI-O-1934724