A SPEECH EMOTION RECOGNITION MODEL GENERATION METHOD USING A MAX-MARGIN FRAMEWORK INCORPORATING A LOSS FUNCTION BASED ON THE WATSON-TELLEGEN'S EMOTION MODEL

摘要

PURPOSE: A method for establishing a model which can recognize feelings in a voice through a loss function and a maximum margin technique based on WTM(Watson-Tellegen Emotional Model) is provided to remarkably increase feeling recognition performance included in a voice. CONSTITUTION: A difference between each emotional feelings is figured by suing geometric distance between emotion groups of WTM(310). Based on set values in the first step, a value of a loss function is obtained(330). Based on a loss function, a parameter of each speech emotion module through a max-margin with margin scaling method is obtained(340).