Voice annotation method for Chinese speech emotion database combined with electroglottography

Abstract

The invention provides a voice annotation method for a Chinese speech emotion database combined with an electroglottography. The main annotation content of the voice annotation method comprises eight layers of information which are simultaneously annotated on each voice. The eight layers of information comprises that a first layer is a text conversion layer, speaking content of a speaker and corresponding paralanguage information are made clear; a second layer is a syllable layer, a regular spell and a tone of each syllable are annotated; a third layer is an initial / final consonant layer, initial / final consonants of the syllable layer are annotated separately, and meanwhile tone information is marked; a fourth layer is an unvoiced sound, voiced sound and silence layer, and unvoiced sounds, voiced sounds and silences of the voices are segmented combined with the electroglottography; a fifth layer is a paralanguage information layer, and paralanguage information included in each voice is annotated; a sixth layer is an emotion layer, and according to emotion status which is expressed by the speaker, each voice is annotated with information comprising seven kinds of emotions and expression degrees of each kind of the emotions; a seventh layer is a stress index layer, and intensity information of pronunciation of each voice is annotated; an eighth layer is a statement function layer, and a statement type of each statement is annotated.

Description

technical field [0001] The invention relates to a method for voice labeling of a Chinese speech emotion database combined with an electroglottogram. The voice label information is more comprehensive and detailed. The electroglottogram is used to directly reflect the vibration information of the vocal cords, thereby reducing noise interference and improving the accuracy of voice labeling. . Background technique [0002] Speech is one of the most direct ways for people to communicate and express their emotions. With the continuous development of human-computer interaction and affective computing, recognizing the emotional information to be expressed from speech has attracted more and more attention from scholars at home and abroad. Corpus is the basis of speech emotion recognition, so it is extremely important to choose a suitable corpus as the basis of speech emotion analysis and recognition. Segmenting and labeling continuous speech plays an important role in making full u...

AI Technical Summary

Problems solved by technology

[0003] In recent years, researchers at home and abroad have established multiple speech emotion databases based on their own scientific research tasks. However, due to the complexity of emotions, there is still no unified standard for the establishment and labeling of emotional speech databases. The database is of great significance to the research of speech emotion recognition and speech synthesis.

Images