The object of this study is to estimate by acoustic methods the emotional trouble of civil aviation pilots. We disposed of the phonic recording machine (CVR) from an aircraft with serious difficulties which finally crashed. We extracted three hundred phonemes from each pilot; we have had these analysed with the aid of the signal processing software I.L.S.
We are inclined towards the idea of a mean spectrum calculated from auto-regressive model coefficients and towards the distribution of energy in this spectrum. For all that, we have calculated cumulative distributions of the sound level in the spectrum. For each phoneme, three cumulative distributions have been processed. One for the band from zero to five kilohertz, the other for the band of the first two formants (0-1875Hz), the third for the complementary band reaching five kilohertz.
We have then been able to show clearly, for the phoneme /a/, a criterion of spectral balance. In a neutral emotional state (when the incident doesn't occur), these cumulative distributions present for the bands 0-1875 Hz and 1875-5000 Hz an important difference. In fact, there is more energy in the low frequencies than in the high frequencies (unbalanced spectrum). In contrast, for the emotionaly charged periods, the spectrum is balanced.
We have thus been able to show that a correlation exists between these objectives measures in the spectrum and the subjective estimations of the emotional charge. However, a study more representative of the statistical point of view, should take place to systemize this report.
Cite as: Gramatica, B., Ruiz, R., Legros, C. (1992) A parameter for the evaluation of emotional trouble: the cumulative distribution of amplitude in the speech spectrum. Proc. ETRW on Speech Processing in Adverse Conditions, 235-238
@inproceedings{gramatica92_spac,
author={B. Gramatica and R. Ruiz and C. Legros},
title={{A parameter for the evaluation of emotional trouble: the cumulative distribution of amplitude in the speech spectrum}},
year=1992,
booktitle={Proc. ETRW on Speech Processing in Adverse Conditions},
pages={235--238}
}