Speaker identification and conclusive evidence system

摘要

(57) Abstract This invention is attenuated and the non audio field component in order to normalize the speech component, regards speaker recognition manner and the system which supply adaptation component summation to each frame of speech. It is used in order to form the new transfer function where all the pole models of linear prediction have the motion even component. The spectrum which is normalized is decided from new transfer function. The spectrum which possesses the property where the speech component is improved and is normalized is limited. From the speech component which is improved, you can obtain the speaker recognition which is improved on the channel.