Research On Speech Emotion Recognition Based On Multiple Feature Combination

Speech is one of the most convenient means of communication between people and it is one of the fundamental methods of conveying emotion as well as semantic information. More over, Emotion plays an important role in communication. So emotion information processing in speech signals has gained increasing attention during the last few years as the need for machines to understand human well in human-machine interaction has grown. Being one of the most branchs of emotion information processing in speech, emotion recognition in speech is the fundamental of the nature human-machine communication. However, the research of emotional recognition still needs further study. The building of emotional speech database, the selection and extraction of emotional characteristic parameters, and the emotional recognition have not formed systematic theory. Therefore, it can be said that speech emotional recognition is still in the preliminary stage, and more deep research is needed.In this work, we firstly overview application the research field involved speech signals emotion recognition, and simply understand the latest research developments of speech emotion recognition in recent years. After analyzing the methods currently used by others, most of the traditional feature extraction extracts only prosodic features reflecting arousal dimension, and quality features is not applied, proposed combination of quality features, MFCC and prosodic features of emotional feature extraction, and uses principal component analysis for extracted original features to reduce the dimension and redundant processing, finally, the speech emotion is recognized by support vector machines.1. An emotional speech database has been build. The article selected 12 emotional text,10 school students were used to reading with different emotions, recorded 4 different emotional speech database with a happy, anger, surprise and sadness, selected 2440 text as the experimental data.2. Analysis and extraction of emotional characteristic parameters. The article extracted parameters form quality features, prosodic features and MFCC features Prosodic features usually are comprised of pitch, amplitude and speed. Quality features usually are comprised of formant, harmonic noise ratio and Mel sub-band energy. The article extracted 12 prosodic features,16 quality features and 12 MFCC features.3. The research of feature selection algorithm and speech emotion recognition. Speech emotion recognition can be viewed as a pattern recognition problem, which is built on the basis of feature extraction. The paper made use of 40 feature parameters, adopt principal component analysis to reduce dimension, then employ SVM to emotion recognition, and analyze the recognition results, verify the validity of voice quality features and MFCC features.