Study on Emotional Speech Recognition to Estimate Speaker’s Emotion and Condition

Abstract

With a goal of developing system by which a user can return to his/her desired scene, we try to recognize emotion using emotional speech. As the first step to recognize emotion, we start to recognize laughing voice for recognizing emotion of pleasure. A previous study start to recognize a laughing voice based on the periodic waveform of the laughing voice. However, a problem arises when recognizing laughing voices through their waveform, since it is possible to have false positive results. To overcome this problem, we proposed laughing voice recognition method that incorporates a voice-likeness feature. We constructed laughing voice recognition system based on method that incorporates a voice-likeness feature. To confirm the efficacy of the laughing voice recognition system, we evaluate this system using laughing voice data and other voice. In our evaluation, we were able to correctly recognize laughing voice at a success rate of 88%.