Breath as a biomarker: A survey of contact and contactless applications and approaches in respiratory monitoring

Contactless breath monitoring technologies like Wi-Fi and acoustic sensing offer practical alternatives to traditional contact-based methods, potentially improving comfort and enabling long-term health monitoring without requiring patients to wear devices or sensors directly on their bodies.

Machine learning and deep learning techniques are increasingly central to breath analysis, with the survey highlighting that different computational approaches are suited to different monitoring scenarios—from detecting individual respiratory rates to identifying diseases—suggesting that one-size-fits-all solutions are insufficient.

Privacy and data security are significant challenges in breath monitoring systems, particularly in multi-user environments where contactless methods could theoretically identify or track multiple people simultaneously, raising concerns about consent and surveillance that developers must address proactively.

Current breath analysis research faces critical limitations including scarce datasets and multi-user interference, which could hinder real-world deployment; the survey emphasizes emerging solutions like federated learning (which trains AI models without centralizing sensitive health data) and transfer learning as promising approaches to overcome these barriers.

Explainable AI is identified as an emerging priority in this field, suggesting that as breath-based health monitoring systems become more sophisticated, there is growing recognition that healthcare providers and patients need to understand how these AI systems reach their conclusions—not just trust their outputs.