Monitor blood glucose levels via breath analysis system and sparse representation approach

It has been reported that the abnormal concentration of acetone in exhaled air is an indicator of diabetes and the concentration rises progressively with the blood glucose level of patients. Therefore, the acetone in human breath can be used to monitor the development of diabetes. In this paper, we introduce a breath analysis system to measure acetone in human breath, and therefore to evaluate the blood glucose levels of diabetics. The system structure, breath collection method, and signal preprocessing method are introduced. To enhance the system performance, we use a novel classification approach, i.e., Sparse Representation based Classification (SRC), to classify diabetics' breath samples into different blood glucose levels. Experimental results show that coupling with SRC, the system is able to classify these levels with satisfactory accuracy.