A TRAINING METHOD TOLERANT OF EMOTIONAL VARIATION FOR SPEECH RECOGNITION SYSTEM

摘要

PURPOSE: A learning method increasing feeling sensitivity in a voice recognition system based on learning data is provided to increase the recognition performance of the voice recognition system when the voice is inputted. CONSTITUTION: A voice parameter extraction module(202) receives learning data emotion and applies frequency conversion parameter. A voice recognition learning module extracts a voice parameter module by inputting a voice parameter. A learning data is converted into the voice parameter by extracting each conversion parameter.