A New Approach for Smoking Event Detection Using a Variational Autoencoder and Neural Decision Forest

Abstract

Smoking is associated with cancer, cardiovascular disease and premature death and can cause severe fire hazards. To assist with smoking cessation, this paper presents a wireless body area networkbased system consisting of two off-the-shelf devices, one smartphone and one smartwatch, to detect smoking events by mining the inertial sensor data from both devices. In the system, an end-to-end trainable unified model is implemented by combining a variational autoencoder with a random forest to classify the collected data into smoking and nonsmoking after data preprocessing. The variational autoencoder is adopted to learn the feature representation and deal with the class imbalance problem, and the stochastic decision forest is adopted to guide the global optimization of the parameter learning process. Extensive validation of our scheme is performed on a large dataset we collected, and our scheme yields quite promising results in terms of accuracy and efficiency.

Enhanced Human Activity Recognition Using Wearable Sensors via a Hybrid Feature Selection Method

“…The data were obtained at 50 samples per second from each phone’s accelerometer, linear acceleration sensor, gyroscope, and magnetometer. Similarly, the SMT dataset [ 35 ] was obtained by recording sensor data at 50 Hz from the three-axis accelerometer and three-axis gyroscope from both a smartphone and wrist-worn device worn by nine participants when performing 12 daily activities. These datasets all have been used and tested in some previous related works and will be utilized for evaluation and analysis here.…”

“…Shoaib et al [ 9 ] used motion sensors from a smartphone at the right trouser pocket and a smartwatch at the right wrist to recognize 13 different activities, and the results showed that the combination of these two positions outperformed either of them alone. When there are more than two devices worn on the body, it is necessary to use a wireless body area network to integrate sensing, computing, and wireless communication into a unified framework, as in studies [ 10 , 35 ].…”

An Efficient ResNetSE Architecture for Smoking Activity Recognition from Smartwatch

“…Recently, increased research has focused on recognizing smoking activities. Video-based and sensorbased approaches are the two most common methods for classifying smoke occurrences and distinguishing human activities [1,3,7,10,11]. The former examines images or videos from a single camera that contains human motion, whereas the latter examines signal data from sensors like gyroscopes, accelerometers, WiFis and sound sensors [12,13].…”

“…Systems that can identify when a person is about to relapse anywhere and at any time are required to help kick the habit. The process of mining statistical data such as smoking recurrence can be used to better comprehend smoking patterns [1] and help people to enhance their standard of health and boost their wellness. Inertial sensors are now utilized to track physical activity utilizing wearable intelligent products such as smartwatches.…”

Are Machine Learning Methods the Future for Smoking Cessation Apps?

“…Typical smartphones are equipped with multiple sensors that are able to collect information such as the user location (Global Positioning System, GPS), identify other people who are near the user (by detecting Bluetooth signals), identifying users’ movements (walking/driving, etc., via onboard accelerometers) and potentially also additional information (such as specific arm movements), if connected to other devices such as smartwatches, which have also shown an accelerating increase in use around the world [ 11 ]. Combining this information with smoking events can enable the characterisation of smoking behaviours to be generated for individual smokers [ 12 ].…”