Non-contact Breathing Patterns Recognition with FMCW Radar by Processing Temporal Information using Transformer Network

This paper proposed non-contact monitoring by classifying breathing patterns using a frequency-modulated continuous wave (FMCW) radar sensor. First, the data were from the subject to conducting breathing pattern: normal, quick, hold, deep, deep quick. Second, we extract the breath wave patterns by employing several block preprocessing and extracting its temporal information that was not investigated by our previous work. Then, to distinguish those five breath wave conditions, the designated Transformer architecture is proposed. Finally, we compared our previous work and demonstrated that our model could enhance accuracy and attained 98.6% of F1-Scores.