A methodology for monitoring emotional stress in phonation

Stress in phonation is mainly shown in the signature of the fundamental frequency. The proposed methodology is based on the estimation of the vocal fold biomechanics in terms of the distribution of the dynamic mass and the mechanical tension of the vocal fold structure. These parameters are derived from the reconstruction of the glottal source by inverse filtering. The vocal fold mechanical tension correlates (stress and strain), are used as the bases for tremor estimation. The correlates of tension and tremor are used to characterize the spontaneous speech of a database of 40 speakers of both genders (20 male and 20 female). Spontaneous speech consists in short interviews of 20 s of duration where the speakers have to express opinions on hot issues with which they are in agreement (pro) or in disagreement (con) following Arciuli's methodology. The emotional stress is estimated from the biomechanical correlates expressed above (tension and tremor). The null hypothesis formulated as the insensitivity of the speaker to pro and con situations has to be disregarded in view of the results for both genders. Interesting open questions are to be raised regarding the possibility of speakers consciously hiding their true opinion based on political correctness. The discussion will offer different hypotheses to further exploit the objective of detecting self-congruence in spoken messages