THE SAME SPEAKER'S VOICE TO CHANGE THE ALGORITHM FOR SPEECH RECOGNITION ERROR RATE REDUCTION

摘要

PURPOSE: An algorithm for reducing a voice recognition error ratio for the voice change of a speaker is provided to confirm a user with a changed voice according to the physical condition of a speaker. CONSTITUTION: Voice data is obtained from a plurality of speakers. The voice data is classified according to the speakers. A voice distribution chart of the speaker is analyzed based on a support vector machine model. A slope is calculated according to the voice distribution chart. The weighted value is calculated according to the voice distribution chart of each speaker.