Research Of Speech Emotion Recognition Based On Text-related

Recently, with the computer plays more and more important role in people’s lives,people’s interest in human-computer interaction (HCI) is increasing and more eager tohave more humane, intelligent and harmonious system, all of this promote theresearch of emotion recognition.Emotion recognition, including Automatic recognition of facial expression,emotion recognition of speech signals, the body movements of emotion recognitionand emotion recognition of physiological signals, etc, based on recent research, withthe aim of reducing time complexity and improving recognition rate, the paper doesresearch as follows. Firstly, give the analysis of emotional, talk about the emotionaldefined by researchers and the establishment of the emotion model to determine thespeech emotion recognition system needs to identify the type of emotional. Secondly,study the speech signal processing, speech signal processing to extract the importantvoice emotional features is an important means of speech emotion recognition, thepaper summarizes the researchers on the characteristic parameters of the speechemotion research, while identifying papers Speech Emotion Recognition System inthe choice of characteristic parameters. Same time, for the complexity of the voice, inspeech emotion recognition, there are not a large, unified voice emotional speechdatabase to reflect the linguistic differences, gender differences, age differences andgeographical differences, this paper gives the method how to create the emotionrecognition library. From the definition and classification of emotions, identify happy,fear, sadness, anger and neutral as this study of five different emotions, then select theSentences which do not contain any emotion themselves, were read with fiveemotions．Then through observing and analyzing, we selected the most important indistinguishing emotions. In the finally, this paper presents the method for speechemotion recognition based on text-based.Before the Emotion Recognition System designed，we trained staff for the voiceemotional database constructed，then extract the fundamental frequency, formant,speed, and energy to construct voice emotional pattern library, at the same time, fromthe view of statistical and the introduction of the weight concept, the method ofmatching has been improved, and finally, the paper chose two sets of experiments,multi-angle verify that the voice emotional and from the text related to text-independent expansion. Experiments show that this speech emotion recognitionsystem with a good speech emotion recognition.