Breath Analytics using high-resolution measurements in everyday settings and their implications to the science of the breathing process

Rifaa Qadri, Faizan Wajid, Mara Cai, Ashok Agrawala

Computer Science Department, University of Maryland, College Park, MD

Poster PDF

Breath analysis holds significant promise for revolutionizing disease detection. While most of the data on breathing has been collected during sleep studies in a clinical setting, we seek to explore the fundamental science of breathing by analyzing high-resolution data collected throughout the day. We utilize Spire Tags which can be worn 24/7 and provide readings of torso movement, accelerometer, and PPG at 25Hz. This enables in-depth analyses of the breathing process and its connections to various activities. We begin by identifying individual breaths and isolating other events such as coughing. Next, we investigate the feature set such as the duration and depth of the inhalation and exhalation phases along with the shape of the curve of each breath. We then look at the sequence of breaths and their characteristics. Our data set was collected from about 80 volunteers who also reported respiratory infections during a period of 2-3 months. Preliminary tests reveal the potential of studying breathing patterns and their ability to detect infections and glean insight into an individual's physical and mental state.