Exploration of glottal characteristics and the vocal folds behavior for the speech under emotion

Abstract

We preliminarily explore physiological characteristics of the vocal folds when the subject is under different emotional modes. Glottal variations from speech production representing the vocal folds behavior will be mainly discussed. We believe that emotion of the human subject has specific impact on the behavior of the vocal folds, which may result in the variations in glottal flow. This paper investigates the physiological characteristics of the vocal folds through variation in the glottal flow under different emotions. A modified algorithm, Pitch Synchronous Iterative Adaptive Inverse Filtering using Average Magnitude Difference Function based on Empirical Mode Decomposition (AMDFEMD-PSIAIF), is proposed to estimate the glottal flow. Glottal flow is discussed and measured by parameters representing the variations in the vibration behavior of the vocal folds. The physical parameters characterizing the muscle tension and viscosity of the vocal folds are estimated using a speech production model, and a fitting method using the glottal flow is proposed. Through an evaluation on a dataset containing over 1200 voice signals, the glottal and physical parameters are measured to verify the variation mode in vocal folds vibration. We obtain the true positive rate and the average sensitivity of each 9 parameter in 6 different emotional modes. Experimental results show that vocal folds present obvious physical changes when under different emotional modes. The CT muscle of the vocal folds is contracting for fear, happy, angry and surprise mode. The TA is relaxing for happy, angry and sad, while contracting when the speaker is under surprise. Fear and surprise make the surface of the vocal folds sticker, while viscosity reduction occurs when the speaker is experiencing sadness. Therefore, the vibration mechanism and physiological properties of the vocal folds corresponding to emotion modes are preliminarily explored.