MoSSe量子ドットを用いたフレキシブルセンサーによる呼吸信号のリアルタイム深層学習分類
A flexible, disposable respiration sensor was developed using Janus-structured MoSSe quantum dots (MoSSe QDs), and its performance was evaluated alongside a deep learning classification framework. The MoSSe QD material demonstrated stable sensing characteristics under humid conditions, with electron affinity and work function analyses indicating a pronounced tendency to interact with hydrogen molecules. A one-dimensional convolutional neural network (1D CNN) was applied to classify four distinct breathing patterns—normal, slow, deep, and fast—achieving 10-trial classification accuracies of 98.18%, 95.25%, 97.64%, and 98.18%, respectively. The results suggest that this low-cost, biocompatible sensor platform holds promise for wearable personal health monitoring applications.
The electron affinity and work function properties of MoSSe quantum dots confer a high tendency to donate electrons to hydrogen molecules, enabling stable respiration sensing even under humid conditions without significant changes in wear rate.
The delivery route is not clearly identifiable from this paper. For hydrogen intake, inhalation is the most efficient route; inhalation, however, carries explosion risk (empirical LFL of 10%; high-concentration devices are not recommended).
See also:
https://h2-papers.org/en/papers/34612334