The brain works in cooperation with the peripheral nervous system (PNS) in maintaining homeostasis and responding to stressful circumstances in daily life. Hypo or hyperactivity of PNS activity to stress has been linked to psychiatric and cardiovascular conditions. How PNS reactivity occurs due to different psychological states is therefore widely studied in daily life settings. An often forgotten confounding variable of PNS activity parameters is speaking.
Speaking changes heart rate and blood pressure more than the changes elicited by psychological stress. If speech goes undetected, periods of speaking can therefore be incorrectly interpreted as periods of stress. Furthermore, speaking “cancels out” the effects of stress on heart rate variability parameters, which can result in the interpretation of stressful speech episodes as relaxation. To combat these issues, we trained and validated speech detection models for different types of respiration and movement (accelerometer) signals. Following a comprehensive machine learning pipeline, the models from these different methods (signals) were statistically compared. The model using upper-sternum mounted movement signals outperformed others, and showed excellent classification performance. We deployed these open-access speech detection models. These can help ambulatory psychophysiology researchers to perform speech detection and control for the confounding effects of speech on physiological parameters.
Download the graphical abstract of this paper using this link.
By highlighting the importance of speech detection and providing multiple validated machine learning models using biosignals to achieve this detection, the results help mitigate the confounding effects of speech in ambulatory psychophysiology research. This ultimately serves to clarify the linkages between perceived psychological states and autonomic reactivity in daily life.
Speech Detection via Respiratory Inductance Plethysmography, Thoracic Impedance, Accelerometers, and Gyroscopes: A Machine Learning-Informed Comparative Study. Saygin, M., Schoenmakers, M., Gevonden, M., & de Geus, E.J.C.
Psychophysiology. 2025 Feb;62(2):e70021.
