摘要 |
PURPOSE: A real time voice recognizing method reducing misrecognition and securing a blocking rate is provided to reduce a generation rate of misrecognition by securing a command recognition blocking rate. CONSTITUTION: A real time voice recognizer sets the recognition of a keyword based on a blocking mode. The recognizer recognizes the keyword in a HMM(Hidden Markov Model) mode(a). When the keyword recognition is successfully performed, the recognizer performs DTW(Dynamic Time Warping) recognition using a stored voice part. The recognizer tries command recognition in the HMM mode in a case of a registered speaker. When the command recognition is successfully performed, the recognizer transmits a protocol for controlling a device corresponding to the recognized command(b). [Reference numerals] (AA) Step (a): keyword recognition; (BB) Keyword voice recognition in an HMM mode; (CC) DTW voice recognition; (DD) Teller candidate and teller registration; (EE) Success of keyword recognition; (FF) Failure of command recognition; (GG) Step (b): command recognition; (HH) Command voice recognition in an HMM mode; (II) Returning to the initial mode in case of fail command recognition; (JJ) Device control, teller voice update, or teller registration in case of succeeding command recognition
|