Figure 4 - uploaded by Casey A Cole
Content may be subject to copyright.
Model of a smoking session: (a) puff duration >0.75 seconds, (b) maximum rest time between puffs <4 minutes and minimum rest time >2.5 seconds, (c) minimum number of puffs in a session=3 puffs, (d) session duration <8 minutes. 

Model of a smoking session: (a) puff duration >0.75 seconds, (b) maximum rest time between puffs <4 minutes and minimum rest time >2.5 seconds, (c) minimum number of puffs in a session=3 puffs, (d) session duration <8 minutes. 

Source publication
Article
Full-text available
Background: Smoking is the leading cause of preventable death in the world today. Ecological research on smoking in context currently relies on self-reported smoking behavior. Emerging smartwatch technology may more objectively measure smoking behavior by automatically detecting smoking sessions using robust machine learning models. Objective: T...

Contexts in source publication

Context 1
... smoking session can be defined in terms of its dependent components such as the number of individual gestures and their time dependencies. Figure 4 describes the model of smoking that was empirically derived based on our observations of the participants' data. Based on this model, a smoking session is described by five main parameters: minimum puff duration, minimum and maximum rest time between puffs, maximum session duration, and the minimum number of puffs per session. ...
Context 2
... "puff" was defined as the time it takes a person to raise the cigarette to their lips, inhale, and then lower their arm back to the resting position. Therefore, we conservatively define a minimum puff duration consisting of 0.75 seconds (shown in Figure 4a). Any puff shorter than 0.75 seconds in duration was therefore rejected as a valid puff by the rule-based AI system. ...
Context 3
... minimum of 2.5 seconds and a maximum of 4 minutes were used as the rest time that separated two adjacent puffs ( Figure 4b) belonging to the same smoking session. Two adjacent puffs in violation of the minimum separation criterion were classified by the rule-based system as the same puff that was incorrectly separated from each other. ...
Context 4
... a smoking session was defined to consist of at least 3 puffs that satisfy the previous gesture criteria (eg, puffs must be longer than 0.75 seconds in duration and more than 2.5 seconds and less than 4 minutes from the next puff) and not exceed 8 minutes in duration (Figure 4c-d). The 8-minute rule was implemented to have a higher precedence over all other rules. ...
Context 5
... this work, we define our objective as successful detection of each smoking session. The interpretation rules of a smoking session ( Figure 4) were used to quantify the output of the smoking detection mechanism. The validity of each detected session was established based on comparison to the self-report by the participants. ...

Similar publications

Preprint
Full-text available
1. The ecological implications of body size extend from the biology of individual organisms to ecosystem-level processes. Measuring body mass for high numbers of invertebrates can be logistically challenging, making length-mass regressions useful for predicting body mass with minimal effort. However, standardised sets of scaling relationships cover...
Article
Full-text available
Sea surface temperature (SST) and salinity (SSS) are essential variables at the ocean and atmosphere interface when considering risk factors for disease in farmed and wild fish stocks. Ecological research has witnessed a recent trend in use of digital and satellite technologies, including remote-sensing tools. We explored spatial coverage of remote...
Article
Full-text available
1. Long-term photographic recordings of animal populations provide unique insights in ecological and evolutionary processes. However, image acquisition at remote locations under harsh climatic conditions is highly challenging. 2. We present a robust, energetically self-sufficient and remote-controlled observatory designed to operate year-round in...

Citations

... In the context of smoking, smartwatches have been demonstrated to be usable for in-situ study of smoking [21], [22] with high accuracy [21]- [24]. Smartwatches have been used to detect smoking gestures with 95% accuracy in a laboratory environment [25] and 90% in-situ detection of smoking [23]. The study of smoking has also been demonstrated to be more accurate when compared to selfreport (90% versus 78%) [23], [26]. ...
... Smartwatches have been used to detect smoking gestures with 95% accuracy in a laboratory environment [25] and 90% in-situ detection of smoking [23]. The study of smoking has also been demonstrated to be more accurate when compared to selfreport (90% versus 78%) [23], [26]. ...
Preprint
Full-text available
The use of sensors available through smart devices has pervaded everyday life in several applications including human activity monitoring, healthcare, and social networks. In this study, we focus on the use of smartwatch accelerometer sensors to recognize eating activity. More specifically, we collected sensor data from 10 participants while consuming pizza. Using this information, and other comparable data available for similar events such as smoking and medication-taking, and dissimilar activities of jogging, we developed a LSTM-ANN architecture that has demonstrated 90% success in identifying individual bites compared to a puff, medication-taking or jogging activities.
... An artificial neural network is a supervised predictive machine learning technique that is successful in classification tasks in various applications [19]- [21]. A feed-forward neural network is a subset of neural networks in general where there are no cycles formed between neurons. ...
Preprint
Full-text available
Sleep studies are imperative to recapitulate phenotypes associated with sleep loss and uncover mechanisms contributing to psychopathology. Most often, investigators manually classify the polysomnography into vigilance states, which is time-consuming, requires extensive training, and is prone to inter-scorer variability. While many works have successfully developed automated vigilance state classifiers based on multiple EEG channels, we aim to produce an automated and open-access classifier that can reliably predict vigilance state based on a single cortical electroencephalogram (EEG) from rodents to minimize the disadvantages that accompany tethering small animals via wires to computer programs. Approximately 427 hours of continuously monitored EEG, electromyogram (EMG), and activity were labeled by a domain expert out of 571 hours of total data. Here we evaluate the performance of various machine learning techniques on classifying 10-second epochs into one of three discrete classes: paradoxical, slow-wave, or wake. Our investigations include Decision Trees, Random Forests, Naive Bayes Classifiers, Logistic Regression Classifiers, and Artificial Neural Networks. These methodologies have achieved accuracies ranging from approximately 74% to approximately 96%. Most notably, the Random Forest and the ANN achieved remarkable accuracies of 95.78% and 93.31%, respectively. Here we have shown the potential of various machine learning classifiers to automatically, accurately, and reliably classify vigilance states based on a single EEG reading and a single EMG reading.
... In order to track cigarette consumption, mobile applications implement two types of loggers. The first type is subjective and requires the smoker to enter the smoking of each cigarette, while the second is objective and rely on smartwatches (Cole et al. 2017;Shoaib et al. 2018) or proprietary wristbands (Lopez-Meyer et al. 2013;Lu et al. 2019) to detect the number of cigarettes and puffs. The Pivot app goes one step further and employs a breath sensor connected to the smoker's mobile device. ...
Article
Full-text available
Despite the association between tobacco use and the harmful effects on general health as well as male fertility parameters, smoking remains globally prevalent. The main content of tobacco smoke is nicotine and its metabolite cotinine. These compounds can pass the blood-testis barrier, which subsequently causes harm of diverse degree to the germ cells. Although controversial, smoking has been shown to cause not only a decrease in sperm motility, sperm concentration, and an increase in abnormal sperm morphology, but also genetic and epigenetic aberrations in spermatozoa. Both animal and human studies have highlighted the occurrence of sperm DNA-strand breaks (fragmentation), genome instability, genetic mutations, and the presence of aneuploids in the germline of animals and men exposed to tobacco smoke. The question to be asked at this point is, if smoking has the potential to cause all these genetic aberrations, what is the extent of damage? Hence, this review aimed to provide evidence that smoking has a mutagenic effect on sperm and how this subsequently affects male fertility. Additionally, the role of tobacco smoke as an aneugen will be explored. We furthermore aim to incorporate the epidemiological aspects of the aforementioned and provide a holistic approach to the topic.
... While it is clear that ML, combined with up-to-date technological advances such as those provided by mobile phones and smartwatches, can provide a perfect platform for delivering targeted personalised and timely interventions [81,82], these methods suffer from the limitation that the smoker needs to have an extra wearable device besides their personal smartphone to predict smoking events. ...
Article
Full-text available
Smoking cessation apps provide efficient, low-cost and accessible support to smokers who are trying to quit smoking. This article focuses on how up-to-date machine learning algorithms, combined with the improvement of mobile phone technology, can enhance our understanding of smoking behaviour and support the development of advanced smoking cessation apps. In particular, we focus on the pros and cons of existing approaches that have been used in the design of smoking cessation apps to date, highlighting the need to improve the performance of these apps by minimizing reliance on self-reporting of environmental conditions (e.g., location), craving status and/or smoking events as a method of data collection. Lastly, we propose that making use of more advanced machine learning methods while enabling the processing of information about the user’s circumstances in real time is likely to result in dramatic improvement in our understanding of smoking behaviour, while also increasing the effectiveness and ease-of-use of smoking cessation apps, by enabling the provision of timely, targeted and personalised intervention.
... A. Wearables in Human Activity Detection Various studies demonstrate how smartwatches have been used to monitor and detect human motions, such as the case of smoking detection [8]- [11], or fall-detection [12]- [15]. Independent reports [16]- [21] also confirm the usability of smartwatches and other smart wearables in the study of complex human motion behaviors such as eating habits, physical activities, and foot motion [16]- [25]. ...
... Uses of Smartwatches include step-counters [22], sleep monitoring [23], diet monitoring [18] as well as general fitness tracking [24]. Smartwatches have demonstrated [25], [26] high accuracy for detecting smoking gestures [8], [25], [26]. Smoking gestures were detected with 95% accuracy in the laboratory environment [27] and 90% accuracy in-situ [8]. ...
... Smartwatches have demonstrated [25], [26] high accuracy for detecting smoking gestures [8], [25], [26]. Smoking gestures were detected with 95% accuracy in the laboratory environment [27] and 90% accuracy in-situ [8]. Studies to identify smoking via gestures has also been demonstrated to be more accurate when compared to self-reporting (90% versus 78%) [8], [28]. ...
Preprint
Full-text available
Poor medication adherence presents serious economic and health problems including compromised treatment effectiveness, medical complications, and loss of billions of dollars in wasted medicine or procedures. Though various interventions have been proposed to address this problem, there is an urgent need to leverage light, smart, and minimally obtrusive technology such as smartwatches to develop user tools to improve medication use and adherence. In this study, we conducted several experiments on medication-taking activities, developed a smartwatch android application to collect the accelerometer hand gesture data from the smartwatch, and conveyed the data collected to a central cloud database. We developed neural networks, then trained the networks on the sensor data to recognize medication and non-medication gestures. With the proposed machine learning algorithm approach, this study was able to achieve average accuracy scores of 97% on the protocol-guided gesture data, and 95% on natural gesture data.
... With their embedded physiological sensors, wrist-worn unobtrusive wearable devices can be used for detecting stress [22], epileptic seizures [23], anxiety [24], emotions [25] and physical activity [26]. Smartphones also have different embedded sensors such as accelerometer, camera, microphone, speaker and photoplethysmography. Accelerometer sensors can be used for smoking event detection [27] and tremor detection for patients with Parkinson's disease [28]. A microphone can be used for detecting snoring and apnea events [29] and a camera sensor can be used for detecting skin cancer [30]. ...
Article
IoT devices generate massive amounts of biomedical data with increased digitalization and development of the state-of-the-art automated clinical data collection systems. When combined with advanced machine learning algorithms, the big data could be useful to improve the health systems for decision-making, diagnosis, and treatment. Mental healthcare is also attracting attention, since most medical problems can be associated with mental states. Affective computing is among the emerging biomedical informatics fields for automatically monitoring a person’s mental state in ambulatory environments by using physiological and physical signals. However, although affective computing applications are promising to improve our daily lives, before analyzing physiological signals, privacy issues and concerns need to be dealt with. Federated learning is a promising candidate for developing high-performance models while preserving the privacy of individuals. It is a privacy protection solution that stores model parameters instead of the data itself and abides by the data protection laws such as EU General Data Protection Regulation (GDPR) and California Consumer Privacy Act (CCPA). We applied federated learning to heart activity data collected with smart bands for stress-level monitoring in different events. We achieved encouraging results for using federated learning in IoT-based wearable biomedical monitoring systems by preserving the privacy of the data.
... These include generic activities such as step counts, sleep detection, and rest periods, while others include more specific activities such as eating [4], drinking [5], managing diabetes [6,7], or smoking [8]. Previous work has established the use of wristworn devices in observing and interpreting smoking behavior in laboratory settings [9][10][11][12][13] and in situ [8,14]. Some of these devices use proprietary sensors [5,[15][16][17], while others use off-the-shelf devices such as smartwatches [8,9,14,18,19]. ...
... Previous work has established the use of wristworn devices in observing and interpreting smoking behavior in laboratory settings [9][10][11][12][13] and in situ [8,14]. Some of these devices use proprietary sensors [5,[15][16][17], while others use off-the-shelf devices such as smartwatches [8,9,14,18,19]. The use of smartwatches in continuous monitoring of human activities and behavior is compelling for several reasons including their availability, decreasing cost, popularity, and the convenience and completeness of data collection. The data connectivity that is afforded by smartwatches adds a critical component to their appeal, allowing for real-time observation and interpretation of activities that can lead to immediate deployment of the appropriate intervention. ...
... The data connectivity that is afforded by smartwatches adds a critical component to their appeal, allowing for real-time observation and interpretation of activities that can lead to immediate deployment of the appropriate intervention. Artificial intelligence (AI)-assisted detection of smoking with smartwatch technology [8][9][10]14,[18][19][20] can reduce the burden of self-reporting by the user and automate notification of the team of research scientists and the caregivers. Despite their potential, the accuracy and resolution of data collected by smartwatches have not been comprehensively explored in comparison to traditional smoking behavior assessments. ...
Article
Full-text available
Background While there have been many technological advances in studying the neurobiological and clinical basis of tobacco use disorder and nicotine addiction, there have been relatively minor advances in technologies for monitoring, characterizing, and intervening to prevent smoking in real time. Better understanding of real-time smoking behavior can be helpful in numerous applications without the burden and recall bias associated with self-report. Objective The goal of this study was to test the validity of using a smartwatch to advance the study of temporal patterns and characteristics of smoking in a controlled laboratory setting prior to its implementation in situ. Specifically, the aim was to compare smoking characteristics recorded by Automated Smoking PerceptIon and REcording (ASPIRE) on a smartwatch with the pocket Clinical Research Support System (CReSS) topography device, using video observation as the gold standard. Methods Adult smokers (N=27) engaged in a video-recorded laboratory smoking task using the pocket CReSS while also wearing a Polar M600 smartwatch. In-house software, ASPIRE, was used to record accelerometer data to identify the duration of puffs and interpuff intervals (IPIs). The recorded sessions from CReSS and ASPIRE were manually annotated to assess smoking topography. Agreement between CReSS-recorded and ASPIRE-recorded smoking behavior was compared. Results ASPIRE produced more consistent number of puffs and IPI durations relative to CReSS, when comparing both methods to visual puff count. In addition, CReSS recordings reported many implausible measurements in the order of milliseconds. After filtering implausible data recorded from CReSS, ASPIRE and CReSS produced consistent results for puff duration (R2=.79) and IPIs (R2=.73). Conclusions Agreement between ASPIRE and other indicators of smoking characteristics was high, suggesting that the use of ASPIRE is a viable method of passively characterizing smoking behavior. Moreover, ASPIRE was more accurate than CReSS for measuring puffs and IPIs. Results from this study provide the foundation for future utilization of ASPIRE to passively and accurately monitor and quantify smoking behavior in situ.
... The experiments showed a smoking detection rate (i.e., recall) of 90% (see Appendix A-Classification Quality Measures). The work described by Cole et al. [4] represents one of the first steps towards the development of methods by employing a commercial smartwatch inertial sensory data instead of a prototypal system of sensors, with the end-goal of integrating the proposed method in a novel real-time intervention for smoking cessation. It presents a Machine Learning system based on the use of an Artificial Neural Network (ANN) able to detect smoking events using only the accelerometer data collected via a smartwatch. ...
... Such a structured and constrained way to record the signals associated to activities affects the interpretation of the very high achieved performances. Hand-to-mouth and chest sensors Research Prototype 20 61% Lopez-Meyer et al. [3] Hand-to-mouth sensor Research Prototype 20 90% Cole et al. [4] App + Smartwatch Research Prototype 10 89% Smokesense [7] Smartwatch Research Prototype 11 89% Research by CWRU [8] App + 2 Armbands Not on the market 10 98% CigFree [9] App + Property Smartband Starting from $69 100 N/A SmokeBeat [10] App + Smartwatch Free 40 80% StopWatch [11] Smartwatch Research Prototype 13 92% Lu et al. [12] Prototypal wrist sensors Research Prototype 5 88% Senyurek et al. [13] Prototypal wrist sensors Research Prototype 35 93% Senyurek et al. [14] Prototypal wrist sensors Research Prototype 35 97% RisQ [15] App + wrist sensors Research Prototype 15 81% ...
Article
Full-text available
Mobile health technologies are being developed for personal lifestyle and medical healthcare support, of which a growing number are designed to assist smokers to quit. The potential impact of these technologies in the fight against smoking addiction and on improving quitting rates must be systematically evaluated. The aim of this report is to identify and appraise the most promising smoking detection and quitting technologies (e.g., smartphone apps, wearable devices) supporting smoking reduction or quitting programs. We searched PubMed and Scopus databases (2008-2019) for studies on mobile health technologies developed to assist smokers to quit using a combination of Medical Subject Headings topics and free text terms. A Google search was also performed to retrieve the most relevant smartphone apps for quitting smoking, considering the average user’s rating and the ranking computed by the search engine algorithms. All included studies were evaluated using consolidated criteria for reporting qualitative research, such as applied methodologies and the performed evaluation protocol. Main outcome measures were usability and effectiveness of smoking detection and quitting technologies supporting smoking reduction or quitting programs. Our search identified 32 smoking detection and quitting technologies (12 smoking detection systems and 20 smoking quitting smartphone apps). Most of the existing apps for quitting smoking require the users to register every smoking event. Moreover, only a restricted group of them have been scientifically evaluated. The works supported by documented experimental evaluation show very high detection scores, however the experimental protocols usually lack in variability (e.g., only right-hand patients, not natural sequence of gestures) and have been conducted with limited numbers of patients as well as under constrained settings quite far from real-life use scenarios. Several recent scientific works show very promising results but, at the same time, present obstacles for the application on real-life daily scenarios.
... Most studies conducted rely on participants to self-report their smoking behavior. In various studies [2], [6], [7], the accuracy of self-reporting has been shown to be no more than ~76%. To bypass the reliance on self-reporting, some studies have been conducted in laboratory-based settings. ...
... It also allows for "in-time" intervention techniques to be developed. Our previous work [6] in this field has indicated that smoking can be detected using only accelerometer data and single-layer artificial neural networks (ANNs) with an accuracy of ~95% in laboratory settings and ~90% in real-world settings [6], [11]. Independent reports [12]- [16] also confirmed the usability of smartwatches in the study of human behavior. ...
... It also allows for "in-time" intervention techniques to be developed. Our previous work [6] in this field has indicated that smoking can be detected using only accelerometer data and single-layer artificial neural networks (ANNs) with an accuracy of ~95% in laboratory settings and ~90% in real-world settings [6], [11]. Independent reports [12]- [16] also confirmed the usability of smartwatches in the study of human behavior. ...
Preprint
The use of sensors has pervaded everyday life in several applications including human activity monitoring, healthcare, and social networks. In this study, we focus on the use of smartwatch sensors to recognize smoking activity. More specifically, we have reformulated the previous work in detection of smoking to include in-context recognition of smoking. Our presented reformulation of the smoking gesture as a state-transition model that consists of the mini-gestures hand-to-lip, hand-on-lip, and hand-off-lip, has demonstrated improvement in detection rates nearing 100% using conventional neural networks. In addition, we have begun the utilization of Long-Short-Term Memory (LSTM) neural networks to allow for in-context detection of gestures with accuracy nearing 97%.
... In previous related works [1]- [10], the image-based and non image-based technologies were used to detect the Manuscript received December 3, 2019; revised March 6, 2020. ...
... In [1,2,3,8,9], the color information of cigarette object or smokes was analyzed, and the smoking condition or smoking behavior were detected effectively. Besides, in [4,5,6,7,10], the sensor-based (i.e. the non image-based) designs were developed to recognize the smoking behavior. To develop the image-based cigarette detector, in general, the wellknown color based designs will be influenced by light disturbance and insufficient light conditions. ...