Voice as a sensitive biomarker for predicting exercise intensity: a modelling study, 2025, Shuyi Zhou et al

Objective: This study investigates the potential of using voice as a sensitive omics marker to predict exercise intensity.


Methods: Ninety-two healthy university students aged 18–25 participated in this cross-sectional study, engaging in physical activities of varying intensities, including the Canadian Agility and Movement Skill Assessment (CAMSA), the Plank test, and the Progressive Aerobic Cardiovascular Endurance Run (PACER). Speech data were collected before, during, and after these activities using professional recording equipment. Acoustic features were extracted using the openSMILE toolkit, focusing on the Geneva Minimalistic Acoustic Parameter Set (GeMAPS) and the Computational Paralinguistics Challenge (ComParE) feature sets. These features were analyzed using statistical models, including support vector machine (SVM), to classify exercise intensity.

Results: Significant variations in speech characteristics, such as speech duration, fundamental frequency (F0), and pause times, were observed across different exercise intensities, with the models achieving high accuracy in distinguishing between exercise states.

Conclusion: These findings suggest that speech analysis can provide a non-invasive, real-time method for monitoring exercise intensity. The study’s implications extend to personalized exercise prescriptions, chronic disease management, and the integration of speech analysis into routine health assessments. This approach promotes better exercise adherence and overall health outcomes, highlighting the potential for innovative health monitoring techniques.
LINK

 

A delayed PEM voice test would prove to be very interesting too.

 

I can think of many confounding variables for patients with illnesses, though.

 

Interesting. Measuring the intensity has been something I've been struggling with. It is not much of an issue for healthy individuals obviously: you can measure heart rate, VO2, etc. For ME/CFS, the typical measures don't work well because the extreme sensitivity to tiny exertion. We need something non-invasive, highly accurate and precise in real time for the measure to be useful enough.

I have some doubt that voice would serve that purpose because talking while exercising can change the voice. The act of measuring changes the measurement, in other words. My voice usually becomes hoarse and labored after biking 20 miles. Last week, I sang all the way. My voice at the end was clear and open. The voice probably gets effected by the exercise position as well. But it would be great if they can measure the intensity with an analysis of some invariant property from a small sample of voice.