Steady Breathing: Researchers Develop Algorithm to Detect Stress
Russian researchers, working with colleagues from Turkey, have developed an algorithm that detects stress by analyzing breathing patterns. They have also built a smartphone app based on the technology. The approach could help monitor the well-being of people whose jobs involve high levels of risk.
The human body interprets stress as a warning signal and immediately mobilizes to protect itself. During these moments, many bodily systems change how they function, including the respiratory system. Breathing may become faster and less regular, while inhalation and exhalation times change, as does the shape of the respiratory waveform.
Researchers from Petrozavodsk State University and Binali Yıldırım University in Erzincan, Turkey, trained a computer algorithm to analyze several breathing characteristics at once. The research was supported by the Russian Science Foundation, and the findings were published in Biomedical Signal Processing and Control.
What the Experiment Revealed
During the experiment, volunteers moved through several conditions: resting, watching entertaining videos, meditating and taking a stress test. For the stress test, participants had to speak before a three-person panel about their strengths and weaknesses, then perform a counting task aloud. Throughout the experiment, a chest-worn sensor recorded their breathing by tracking chest movements.
The recorded signals were segmented into one-minute intervals and processed using machine-learning algorithms. The researchers compared two approaches. In the first, a compact neural network analyzed the respiratory waveform directly. In the second, the team first extracted features from the signal, including waveform shape, changes in breathing rhythm across different frequencies and temporal stability. Those features were then used to assess the participant’s state.
The neural network that processed the raw signal delivered the best results. Rather than examining individual metrics in isolation, it analyzed breathing as a whole, including inhalation duration, rhythm regularity and waveform shape. With this approach, the AI correctly identified the presence or absence of stress in more than 96% of cases.
A Smartphone Instead of a Sensor
Building on the research, the team developed Breath Couple, an Android app. In the laboratory experiment, breathing was recorded with a chest-worn sensor, but the app uses a different method. It captures inhalations and exhalations through a small external microphone connected to a smartphone.
The microphone must be positioned near the nose. The app then reconstructs the respiratory waveform and allows users to track breathing rate and fluctuations in rhythm. This approach could make monitoring more accessible because users would not need to wear a chest sensor continuously.
For now, the app is being used for personal monitoring and pilot studies. The next step is to test the method with a larger group of volunteers and under everyday conditions. The researchers plan to investigate how movement, talking, body position, desk work, rest and sleep affect the results. These trials should help establish how reliably the algorithm detects changes in breathing outside the laboratory.
Where the Technology Could Be Used
One promising application is in occupations that involve high levels of responsibility and work in extreme conditions. These include drivers, pilots, air traffic controllers and operators of complex industrial systems. If further testing confirms the method’s reliability, breathing analysis could provide an additional way to monitor employees’ condition and identify signs of fatigue, emotional strain and physical overload.
The technology also has potential applications in healthcare. Objective breathing data could support research into stress, anxiety and depression, as well as studies of methods for regulating emotional responses. In sports and education, this type of monitoring could help researchers investigate how physical or mental demands and nervousness affect a person’s physiological state.
Export Potential
Technologies for monitoring people’s condition are in growing demand. The development from Petrozavodsk State University aligns with this global trend while making such monitoring more accessible. Before starting work, an employee could check their condition using just a smartphone and a compact external microphone. If further trials confirm that the method is effective in real-world settings, the algorithm could attract international partners as a ready-to-deploy software solution.
