Study on Breath Gas Information Sensing and Its Application to Man-Machine Interface

提出了一种利用呼吸气体信息实现情绪识别的智能气体传感系统。采用石英晶体谐振器和等离子体聚合物薄膜作为传感器，设计了一种呼吸气体传感系统。针对情感识别的计算实验，研究了神经网络和支持向量机等基于机器学习的方法。在情绪识别实验中，通过使用收集的呼吸气体数据下的心理实验，得到的最大平均情绪识别率为75%，使用人工神经网络和83%，使用SVM的三种情绪：高兴，不高兴，没有情绪。在情绪识别实验中，通过使用漱口水作为刺激的兴奋情绪，得到的最大平均情绪识别率为47.5%，使用人工神经网络的两种情绪：舒适和没有情绪。实验结果表明，使用呼吸气体信息是可行的，基于机器学习的方法是非常适合这项任务。

This paper proposes a smart gas sensing system to achieve emotion recognition using breath gas information. A breath gas sensing system is designed by using s quartz crystal resonator with a plasma-polymer film as a sensor. For computational experiment of emotion recognition, the machine learning-based approaches, such as neural network (NN) and support vector machine (SVM), are investigated. In emotion recognition experiments by using gathered breath gas data under psychological experiments, the obtained maximum average emotion recognition rates are 75% using ANN and 83% using SVM for three emotions: pleasure, displeasure, and no emotion. In emotion recognition experiments by using a dental rinse as a stimulus for exciting emotion, the obtained maximum average emotion recognition rate is 47.5% using ANN for two emotions: comfortableness and no emotion. Experimental results show that using breath gas information is feasible and the machine learning-based approach is well suited for this task.