Medidas objetivas de fadiga e sonolência baseadas na análise acústica da organização temporal da fala

• Ano de defesa: 2023
• Autor(a) principal: Deborah Christiny Abrante Godinho
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; Brasil; ICB - INSTITUTO DE CIÊNCIAS BIOLOGICAS; Curso de Especialização em Neurociências; UFMG
• Palavras-chave em Português: Fadiga mental; Sonolência; Fala; Acústica da fala; Medida da produção da fala; Neurociências; Acústica da Fala
• Link de acesso: http://hdl.handle.net/1843/64559

INTRODUCTION: Fatigue and sleepiness are states that negatively affect cognitive and physical capabilities, with direct implications for aviation safety. Fatigue is often underestimated as a contributing factor in aeronautical occurrences, despite its significant influence on decision-making. Objective methods for measuring fatigue are necessary to better associate it with critical events. Acoustic analysis of voice and speech has proven to be a reliable and robust alternative for detecting a range of individual alterations and expressions, including fatigue and sleepiness, demonstrating that variations in human speech can also be robust indicators of these temporary states. This study aimed to apply an objective method of acoustic analysis based on the temporal organization of speech, (proposed in Vasconcelos (2019), “Fatigue and sleepiness in aviators: analysis of voice, speech and language variations”, doctoral thesis, UFMG, in Portuguese), to a larger sample size in order to confirm and expand upon the obtained results. Additionally, it sought to investigate the correlation of this objective method with subjective methods of fatigue and drowsiness detection. METHODS: A retrospective cross-sectional analysis of a database containing spontaneous speech samples, sleep history data from the last 72 hours, and psychometric scales of fatigue (Samn-Perelli) and sleepiness (Karolinska) from 25 aircraft pilots, native Brazilian Portuguese speakers, and neurologically healthy. Data were recorded on a day off when pilots felt rested and on a workday after being awake for 16 hours. After extracting measures of speech temporal organization, data analysis was performed using paired Student's t-test and correlations through Spearman's correlation test, adopting a 5% significance level. RESULTS: Measures of speech temporal organization showed statistically significant variations when comparing rest and workdays, evidencing a worsening of the analyzed parameters when individuals reported fatigue and sleepiness. The speech rate reduction and the frequency variation of pause types were particularly relevant among the speech measures. A significant correlation was found between speech analysis measures and the Karolinska and Samn-Perelli psychometric scales. Furthermore, multiple linear regression showed that the percentage of pauses is predictive of the level of fatigue and sleepiness reported in these scales. It was found that there exists a high correlation (r = 0.98) between the speech rate and the Samn-Perelli scale, represented by the means of four bands. This result allows the level of fatigue to be inferred using an objective measure (speech rate) instead of a subjective measure (Samn-Perelli scale). CONCLUSIONS: The high variability of the temporal organization parameters of speech in response to the presence of fatigue and sleepiness suggests that these are good indicators that reflect the level of cognitive planning of discourse and the individual's state of alertness. In addition, the methods for extracting speech parameters replicated in this study proved to be appropriate for the objective detection of fatigue and sleepiness.