Research On Rapid Detection Method Of Smoking Based On Smartwatch

Nowadays,as the development of research on somatology and behavioristic,the research projects grow increasingly focusing on the detection of human behavior,in order to provide solutions for human health inspection.Smoking behavior detection,one of the most popular fields,is a promising topic in research.The number of smokers grows annually in the world as well as the cases with respect to harms resulting from smoking.No matter in the aspect of personal health or the safety of public property,the loses are immeasurable caused by smoking.Thus,a solution is especially significant for fast and effective smoking detection that is capable of detecting smoking behavior in the early stage of smoking and providing alarm service so as to increase the probability of smoking cessation,effectually reducing the hazards brought about from smoking.In the current research on smoking detection,one of the solutions detects the smoking behavior by computer vision technique,the application of which is limited tremendously due to detection environment and personal privacy.In addition,some research proves that the strength variation of wireless signal can be utilized for smoking behavior detection.Although it is not susceptible to environment illumination and privacy issues,multipath effect is non-negligible and it is not able to distinguish the behaviors in the level of fine grit,such as smoking and eating with similar acting tracks.Also,with the growing application in wearable devices,a smart watch can be used to detect smoking behavior by analyzing data extracted from inertia sensors,which is regarded as a better option to help smokers to quit smoking because it is an active behavior from the smoker's point of view.Nevertheless,the most fundamental hypothesis for this solution is that the wrist wearing the smart watch and the hand for smoking must be ipsilateral.Therefore,the detection accuracy of this method is not guaranteed.To break the limitations mentioned above,this thesis presents a smoking detection scheme that does not require distinguishing which hand is used,and issues,like the power consumption for all day real time inspection,how to achieve fast and precise detection and the availability for various groups of people in diverse application environment,is addressed.By doing research in the early stage on human physiology and smoking behavior,a unique physiological phenomenon caused by smoking has been discovered.The nicotine,one ofthe ingredients of cigarette,can result in special variation in heart rate,which generates a pattern of square wave that is totally different from the variation induced by other daily life activity.Based on this discovery,the HSM(Heart Smoke Match)algorithm is proposed to match the heart rate patterns of smokers to the one recognized as smoking-induced pattern for smoking detection.This algorithm is the first real-time smoking detection solution based on the heart rate measurement by a smart watch,which is available for smoking detection on either hand and immune to other similar behaviors in a fast and effective manner.Verified by experimental results,the HSM algorithm can detect smoking behavior fast and accurately in different real life environment and its accuracy and recall-rate are 96.7% and 99.8%,respectively.