Emotion Recognition in Mandarin Speech and Its Application in Training of Hearing Impaired

Abstract

Language is the most basic and main tool for the human to communicate thoughts, convey mes-sages and express aspiration. Language learning mainly relies on the coordination and training of the auditory system and articulatory system. For the hearing-normal person, this kind of learning process is very natural. But for the hear-ing-impaired person, it becomes almost impossi-ble without the auditory input. Computer-assisted speech training of hear-ing impaired is a training system to help the hearing-impaired person to speak. In this paper, we propose a Mandarin speech emotion recogni-tion system. In this system, the features we ex-tracted include 16 LPC coefficients and 20 MFCC (Mel Frequency Cepstrum Coefficients). We adopt two common statistical pattern classi-fication methods, the minimum-distance method and the nearest class mean method, to classify the speech into one of the five basic emotions. In the minimum-distance method, the correct rec-ognition rate is 79.1%. In the nearest class mean method, the correct recognition rate reaches 89.1%. By applying this system, it can assist the hearing-impaired people to learn not only to speak correctly but also to speak “naturally”, just like the hearing-normal people.