Recognizing the Emotional State Changes in Human Utterance by a Learning Statistical Method based on Gaussian Mixture Model

Speech is one of the most opulent and instant methods to express emotional characteristics of human beings, which conveys the cognitive and semantic concepts among humans. In this study, a statistical-based method for emotional recognition of speech signals is proposed, and a learning approach is introduced, which is based on the statistical model to classify internal feelings of the utterance. This approach analyzes and tracks the emotional state changes trend of speaker during the speech. The proposed method classifies utterance emotions in six standard classes including, boredom, fear, anger, neutral, disgust and sadness. For this purpose, it is applied the renowned speech corpus database, EmoDB, for training phase of the proposed approach. In this process, once the pre-processing tasks are done, the meaningful speech patterns and attributes are extracted by MFCC method, and meticulously selected by SFS method. Then, a statistical classification approach is called and altered to employ as a part of the method. This approach is entitled as the LGMM, which is used to categorize obtained features. Aftermath, with the help of the classification results, it is illustrated the emotional states changes trend to reveal speaker feelings. The proposed model also has been compared with some recent models of emotional speech classification, in which have been used similar methods and materials. Experimental results show an admissible overall recognition rate and stability in classifying the uttered speech in six emotional states, and also the proposed algorithm outperforms the other similar models in classification accuracy rates.

Keywords:

speech processing , emotional states , Pattern Recognition , mel frequency cepstral coefficient , Gaussian mixture model