Drink-PulseNet: Recognize Drinking Status with Pulse Wave-Based Learning Method

Alcohol restrictions are currently in place in certain specific situations to ensure safety and compliance. Existing alcohol testing methods include breathalyzer, blood and saliva testing, but all have obvious drawbacks. In this paper, we propose a pulse wave-based learning method Drink-PulseNet to recognize the drinking status. The pulse contains a large amount of physiological information. By extracting the time-frequency domain features with neural network, we can determine whether humans are in the state of non-drinking, micro-drinking or drunkenness. Testing results show that the prediction accuracy of our proposed method achieves 88.6%, which can outperform other methods and support the practical applications.