Introduction
Psychomotor fatigue and cognitive load may affect human performance and speech production. Voice analysis offers a potentially non-invasive approach to monitoring such changes. However, review studies have highlighted inconsistencies in findings concerning acoustic indicators of cognitive load [1–3]. This study aimed to preliminarily assess the usefulness of temporal speech parameters and relative voice signal level as complementary measures to conventional acoustic features.
Methods
Thirty recordings from ten participants (six women and four men; aged 20–26 years) were analyzed. For each participant, recordings of the same sentence read aloud were obtained at three stages: baseline, after the Stroop test, and after the Digit Symbol Substitution Test (DSST). The recordings were acquired as 16-bit WAV files at a sampling rate of 16 kHz. Voice activity detection (VAD) based on energy thresholding was implemented in Python. Utterance duration, the proportion of internal pauses lasting at least 0.20 s, and the root mean square (RMS) level of active speech expressed in dBFS were calculated. The analysis was descriptive, without statistical significance testing.
Results
Mean utterance duration was 7.30 s at baseline, 7.62 s after the Stroop test, and 6.80 s after the DSST. The corresponding proportions of pauses were 6.08%, 6.30%, and 2.18%, respectively. Following the DSST, no pauses meeting the predefined criterion were detected in four participants. Mean RMS levels were −34.57, −35.19, and −35.57 dBFS, respectively.
Conclusions
These preliminary findings demonstrate the technical feasibility of automated analysis of short speech recordings and indicate variation in temporal speech parameters across measurement stages. However, the small sample size, fixed task order, and potential effects of sentence repetition prevent attributing the observed changes solely to psychomotor fatigue. Further measurements and improved control of recording conditions are required.
References
[1]SchewskiL.etal.(2025).PLOSONE,20(7),e0328833.doi:10.1371/journal.pone.0328833.
[2]PyfromM.etal.(2026).JournalofVoice,40(1),218–228.doi:10.1016/j.jvoice.2023.08.024.
[3]MacPhersonM.K.etal.(2017).JournalofVoice,31(4),504.e1–504.e9.doi:10.1016/j.jvoice.2016.10.021.