Recognition and Analysis of Emotion Types from Myanmar Movies

Recognizing emotion from speech has become the active research themes in speech processing and in applications based on human-computer interaction. The emotion recognition system is also intended to apply in criminal cases; one potential application is the detection of the emotional state in Automatic Teller Machine (ATM) machine to provide feedback to machine. Especially, if it detect fear emotion, it provide feedback to stop all of the money transition. In this domain, the image resolution of facial expressions captured by CCTV camera can be degraded at night due to the inadquate light. To overcome this problem, this paper conducts an experimental study on recognizing emotions from human speech. The emotions considered for the experiments include angry, disgust, happy, fear, sad and surprise. Feature selection method, which aims to extract effective acoustics features to improve the performance of emotion recognition is focused in this paper. The combination of MFCC (Mels Frequency Cepstral Coefficients) with Berouti Spectral Subtraction is used to obtain compressed feature vectors without losing much information. The standard SAVEES dataset and our self-created Myanmar emotional dataset are used for training and testing. After that, those emotions are identified with Random Forest classifier. Finally, the detail analysis results on enhanced features and only MFCCs features corresponding to different databases and different classifiers are compared and explained.