Use of automated speech analysis and facial emotion measurements on videos to assess the effects of relaxation devices: a pilot study

Guyon, A., Guerchouche, R., Konig, A., Linz, N., Zimmer, C., Do Carmo Blanco, N., … & Elleuch, E. (2020). Use of automated speech analysis and facial emotion measurements on videos to assess the effects of relaxation devices: a pilot study. Journal of Interdisciplinary Methodologies and Issues in Sciences.

Abstract

Rapid relaxation installations in order to reduce stress appear more and more in public or work places. However, the effects of such devices on physiological and psychological parameters have not been scientifically tested yet. This pilot study (N=40) evaluates the variations of vocal speech and facial emotions parameters in 3-minute videos of participant recorded just before and after relaxation, on four different groups, three of them using a different rapid (15 minutes) sensorial immersion relaxation devices and a control group using no device. Vocal speech parameters included sound duration, pause mean duration, sound duration ratio, mean vocal frequency (F0), standard deviation of F0, minimum and maximum of F0, jitter and shimmer. Facial emotion analysis included neutral, happy, sad, surprised, angry, disgusted, scared, contempt, valence and arousal. The objective of this study is to evaluate different parameters of the automated vocal and facial emotions analysis that could be of use to evaluate the relaxation effect of different devices and to measure their variations in the different experimental groups.
We identified significant parameters that can be of use for evaluating rapid relaxation devices, particularly voice prosody and minimum vocal frequency, and some facial emotion such as happy, sad, the valence and arousal. Those parameters allowed us to discriminate distinct effects of the different devices used: in G1 (control) and G2 (spatialized sounds), we observed a slow-down in voice prosody; in G3 (Be-breathe) a decrease in minimum vocal frequency and an increase of arousal; while in G4 (3D-video) we found an increase in facial emotion valence (happy increasing and sad decreasing). Other parameters tested were not affected by relaxation.