Smoking Activity Recognition Using a Single Wrist IMU and Deep Learning Light

Abstract

Smoking has a strongly relation with diseases such as lung cancer, chronic obstructive pulmonary disease, and coronary heart disease. To prevent smoking, there are various passive ways including warning stickers and electronic cigarettes. However, a smart and proactive methodology might be more effective and useful to break the smoking habit by automatically and actively providing feedbacks to smokers to promote their desire of quitting smoking. In this work, we propose such a smart and proactive system using a wrist band housing a single Inertial Measurement Unit (IMU) sensor, and a smartphone App. housing artificial intelligence based on Recurrent Neural Network (RNN). To detect the smoking puffs, the proposed system uses a two steps classification scheme: first, a General model categorizes measured activities into Activities Daily Living (ADL) and Hand Gestures Activity (HGA). Then an Expert model further categorizes HGAs into smoking, eating, and drinking. Our smoking activity recognition system recognizes smoking activity with an accuracy of 91.38% and provides an active vibration feedback to smokers.