Breathing is automatic, but the way you do it reveals a lot about your state of mind. A new study published in *Technologies* has trained an artificial intelligence to distinguish between stress, calm, and meditation by analyzing breath alone. The result? 92.33% accuracy on the WESAD dataset, a benchmark for wearable physiological signal research.
Researchers extracted 16 parameters from the respiratory signal: rhythm, depth, and nonlinear characteristics. They didn't just feed data into a black box. They used an "ensemble" model (a team of algorithms voting together) and, crucially, made it explainable using SHAP, a tool that shows *why* the AI makes a specific decision.
What emerged? Respiratory rate, inspiration-to-expiration time ratio, and expiratory variability are the main red flags. These parameters mirror autonomic nervous system activity: when stressed, the sympathetic system (the accelerator) overrides the parasympathetic (the brake).
There's a surprise: under stress, breath depth *increases*, contrary to the expectation of short, shallow breaths. This paradox opens new research avenues. The ultimate goal? Smartwatches and wearables capable of monitoring mental health in real-time, alerting you before the load becomes too heavy. A digital compass to navigate the most intense days.
