Search

In many disaster situations, the location of victims is the most important information for saving them. To provide such information, most studies simply use indoor positioning technologies of the victim’s mobile device. Nevertheless, their schemes have a large lack to point out the exact location of a victim because people sometimes drop off their mobile device or put it down some places, such as a desk and a table. In other words, it merely provides the location of mobile devices and not of victims. Hence, this paper proposes a novel virtual sensing mechanism, named the USING sensor (User-Smartphone-Decoupling Detection). The USING sensor relies on the user activity sensing and co-existence monitoring of users and smart mobile devices. It could track the continuous coupling state of the user-mobile device. However, continuous user activity detection is not trivial, since some actions, such as sitting, dropping, and so on, intermittently occur while some actions like walking are frequent and continuous. So, data acquisition methods from a mobile device for the detection of user actions should be designed and managed differently. In addition, for high sensing reliability, the USING sensor is based on machine learning (ML) analytics with multiple algorithms and the ensemble technique. Thus, it could achieve 95.0% of user activity recognition accuracy and 86.8% of sensing accuracy.

Loneliness among college students is an increasingly prevalent issue. While technology-based methods for detection using behavioural patterns have been proposed, there remains an opportunity for improvement as insufficient attention has been given to the individual behavioral differences among students. Loneliness is a highly subjective experience and people’s routines differ, making it a challenge for generic models to accurately determine its presence and severity. In response to this challenge, it is particularly helpful to identify subgroups within the population that exhibit similar behavioral characteristics, enabling a more nuanced understanding and detection of loneliness. This paper introduces a novel approach to loneliness detection, leveraging a data set gathered through passive sensing using mobile phones, which provides a rich source of user behavioral data. We utilized unsupervised clustering to find subgroups of students exhibiting similar behavioral patterns over time within the data set. This approach is essential for continuous monitoring, identifying changes in behavioral patterns, and facilitating the early detection of loneliness. Using data from 41 students’ smartphones, we created group-specific classification models to identify loneliness. Group-based prediction models for loneliness detection have shown significant improvement in accuracy over generalized models. These findings can lead to the development of more effective, tailored methods for loneliness detection in diverse populations. This study emphasizes the importance of personalized approaches in mental health interventions and highlights the potential of passive sensing data in creating tailored loneliness detection methods.

Carbon monoxide (CO) and Nitrogen dioxide (NO2) are major air pollutants that have the potential to affect human health adversely. There is a lack of useful information regarding the spatial distribution and temporal variability of CO and NO2 emissions in major metropolitan areas. The primary goal of this research is to provide a geospatial data methodology for detecting emission spikes of CO and NO2 in polluted urban environments employing portable, low-cost sensors. We propose that ephemeral identification of harmful gas concentrations can be achieved using different IoT device types mounted on a mobile platform. We propose that persistent CO and NO2 emission spikes can be identified by driving through the city on different days. We applied this approach to Hyderabad, India, by fixing a mobile platform on a street car. We corrected the IoT device measurement errors by calibrating the sensing component data against a reference instrument co-located on the mobile platform. We identified that random forest regression was the most suitable technique to reduce the variability between the IoT devices due to heterogeneity in the mobile sensing datasets. The spatial variability of CO and NO2 harmful emission spikes at a resolution of 50 m were identified, but their intensity changes on a daily basis according to meteorological conditions. The temporal variability shows a weak correlation between CO and NO2 concentrations. The data from the CO and NO2 emission spikes at Points of Interest that disturb traffic flows clearly show the need for public education about when it is hazardous for persons with respiratory conditions to be outside, as well as when it is unsafe for young children and the elderly to be outside for extended periods of time. This detection strategy is adaptable to any mobile platform used by individuals traveling by foot, bicycle, drone, or robot in any metropolis.

While the task of automatically detecting eating events has been examined in prior work using various wearable devices, the use of smartphones as standalone devices to infer eating events remains an open issue. This paper proposes a framework that infers eating vs. non-eating events from passive smartphone sensing and evaluates it on a dataset of 58 college students. First, we show that time of the day and features from modalities such as screen usage, accelerometer, app usage, and location are indicative of eating and non-eating events. Then, we show that eating events can be inferred with an AUROC (area under the receiver operating characteristics curve) of 0.65 using subject-independent machine learning models, which can be further improved up to 0.81 for subject-dependent and 0.81 for hybrid models using personalization techniques. Moreover, we show that users have different behavioral and contextual routines around eating episodes requiring specific feature groups to train fully personalized models. These findings are of potential value for future mobile food diary apps that are context-aware by enabling scalable sensing-based eating studies using only smartphones; detecting under-reported eating events, thus increasing data quality in self report-based studies; providing functionality to track food consumption and generate reminders for on- time collection of food diaries; and supporting mobile interventions towards healthy eating practices.

Over the years, mobile phones have become versatile devices with a multitude of capabilities due to the plethora of embedded sensors that enable them to capture rich data unobtrusively. In a world where people are more conscious regarding their health and well-being, the pervasiveness of smartphones has enabled researchers to build apps that assist people to live healthier lifestyles, and to diagnose and monitor various health conditions. Motivated by the high smartphone coverage among young adults and the unique issues they face, in this review paper, we focus on studies that have used smartphone sensing for the well-being of young adults. We analyze existing work in the domain from two perspectives, namely Data Perspective and System Perspective. For both these perspectives, we propose taxonomies motivated from human science literature, which enable to identify important study areas. Furthermore, we emphasize the importance of diversity-awareness in smartphone sensing, and provide insights and future directions for researchers in ubiquitous and mobile computing, and especially to new researchers who want to understand the basics of smartphone sensing research targeting the well-being of young adults.

Authenticating a user in the right way is essential to IT systems, where the risks are becoming more and more complex. Especially in the mobile world, banking applications are among the most delicate systems requiring strict rules and regulations. Existing approaches often require point-of-entry authentication accompanied by a one-time password as a second-factor authentication. However, this requires active participation of the user and there is continuous authentication during a session. In this paper, we investigate whether it is possible to continuously authenticate users via behavioral biometrics with a certain performance on a mobile banking application. A currently used mobile banking application in Turkey is chosen as the case, and we developed a continuous authentication scheme, named DAKOTA, on top of this application. The DAKOTA system records data from the touch screen and the motion sensors on the phone to monitor and model the user's behavioral patterns. Forty-five participants completed the predefined banking transactions. This data is used to train seven different classification algorithms. The results reveal that binary-SVM with RBF kernel reaches the lowest error scores, 3.5% equal error rate (EER). Using the end-to-end DAKOTA system, we investigate the performance in real-time, both in terms of authentication accuracy and resource usage. We show that it does not bring extra overhead in terms of power and memory usage compared to the original banking application and we can achieve a 90% true positive recognition rate, on average.

Vehicle maneuverings are important pieces of information for many applications such as traffic incident detection and driving behavior recognition. On a macroscopic level, an unusually large number of some maneuvering events (e.g., lane changes) may suggest that a road incident occurs. On a microscopic level, vehicle maneuverings can tell how each individual person drives (e.g., safely or unsafely). A number of smartphone-based maneuvering detection methods have been proposed. However, most of them typically rely on accelerometer and gyroscope, and thus require the smartphone to be placed at a specific position on a vehicle and in a specific orientation. In this work, we take a rather different approach from most of the existing methods. Particularly, we investigate how effective it is to detect vehicle maneuverings by relying only on a signal from a digital compass on a smartphone. To this end, we introduce a simple rule-based maneuvering detection method which only takes the heading angles measured by the digital compass as inputs. This allows the smartphone to be placed freely at any position on the vehicle and in any orientation. Our results show that the digital compass can be used to detect turns and u-turns extremely well, and its accuracy on lane change detection is at an acceptable level.

Lane change is a valuable piece of traffic information. An abnormally high number of lane changes on a road section typically suggests that some lanes are blocked due to traffic incidents. Currently, the lane change information of vehicles on an urban road is typically obtained from over-roadway fixed sensors such as surveillance cameras. However, using fixed sensors has limitations in terms of cost and coverage. It is more effective to collect the lane change information directly from each individual vehicle. In addition, lane changing behavior of a driver can help assess his driving risk. One convenient way to detect a lane change directly from each vehicle is to take advantage of sensors on smart mobile devices such as smartphones. In this paper, we explore a new way to identify a lane change event based on a pattern of steering wheel angles detected by a smartphone. This is distinguishable from all of the existing steering wheel-based lane change detection methods, which need to retrieve the steering wheel angle signal from the On-board Diagnostic (OBD) port of the vehicle via the Controller Area Network-Bus (CAN-Bus). In addition, unlike others, our method does not require a lot of complex features to make an accurate detection. In fact, we demonstrate that a high level of accuracy can already be attained with a single simple feature called rotation span. Results show that the proposed detection method performs very well both in terms of precision and recall.

Smartphone magnetometer readings exhibit high linear correlation when two phones coexist within a short distance. Thus, the detected coexistence can serve as a proxy for close human contact events, and one can conceive using it as a possible automatic tool to modernize the contact tracing in infectious disease epidemics. This paper investigates how good a diagnostic test it would be, by evaluating the discriminative and predictive power of the smartphone magnetometer-based contact detection in multiple measures. Based on the sensitivity, specificity, likelihood ratios, and diagnostic odds ratios, we find that the decision made by the smartphone magnetometer-based test can be accurate in telling contacts from no contacts. Furthermore, through the evaluation process, we determine the appropriate range of compared trace segment sizes and the correlation cutoff values that we should use in such diagnostic tests.

Transportation and locomotion mode recognition from multimodal smartphone sensors is useful for providing just-in-time context-aware assistance. However, the field is currently held back by the lack of standardized datasets, recognition tasks, and evaluation criteria. Currently, the recognition methods are often tested on the ad hoc datasets acquired for one-off recognition problems and with different choices of sensors. This prevents a systematic comparative evaluation of methods within and across research groups. Our goal is to address these issues by: 1) introducing a publicly available, large-scale dataset for transportation and locomotion mode recognition from multimodal smartphone sensors; 2) suggesting 12 reference recognition scenarios, which are a superset of the tasks we identified in the related work; 3) suggesting relevant combinations of sensors to use based on energy considerations among accelerometer, gyroscope, magnetometer, and global positioning system sensors; and 4) defining precise evaluation criteria, including training and testing sets, evaluation measures, and user-independent and sensor-placement independent evaluations. Based on this, we report a systematic study of the relevance of statistical and frequency features based on the information theoretical criteria to inform recognition systems. We then systematically report the reference performance obtained on all the identified recognition scenarios using a machine-learning recognition pipeline. The extent of this analysis and the clear definition of the recognition tasks enable future researchers to evaluate their own methods in a comparable manner, thus contributing to further advances in the field. The dataset and the code are available online. http://www.shl-dataset.org/.

With the rapid development of mobile Internet techniques, using the sensor-rich smartphones to sense various contexts attracts much attention, such as transportation mode recognition. The transportation mode information can help to improve urban planning, traffic management and journey planning. Though much work has been done on the transportation mode recognition using classic machine learning algorithms, the performance of these methods is not reasonable and heavily relies on the effectiveness of handcrafted features. In this paper, we leverage the strong representation ability of deep learning method and present a deep-learning-based algorithm for transportation mode recognition, namely CL-TRANSMODE, which is capable of accurately detecting multiple transportation modes. The algorithm first uses a convolutional neural network (CNN) to learn appropriate and robust feature representations for transportation modes recognition. Then, an LSTM network performs a further learning of the temporal dependencies characteristics on the feature vectors of CNN output. To further enhance the accuracy of transportation mode recognition, several artificial segments and peak features are extracted from the raw sensor measurements. These features characterize the transportation modes over a much long period of time (minutes or hours). By combining the CNN-extracted features and handcrafted features, our proposed CL-TRANSMODE transportation mode recognition algorithm can accurately differentiate eight transportation modes, i.e., walking, running, bicycling, driving a car, riding a bus, taking a metro, taking a train, or being stationary. Extensive experiments on both the SHL and HTC datasets demonstrate that use our proposed CL-TRANSMODE transportation mode recognition algorithm which outperforms the state-of-the-art comparative algorithms. On the SHL dataset, which contain barometric data, the accuracy using the CL-TRANSMODE algorithm can reaches 98.1%.

With the rapid development of deep convolutional neural networks, more and more computer vision tasks have been well resolved. These convolutional neural network solutions rely heavily on the performance of the hardware. However, due to privacy issues or the network instability, we need to run convolutional neural networks on embedded platforms. Critical challenges will be raised by limited hardware resources on the embedded platform. In this paper, we design and implement an embedded inference framework to accelerate the inference of the convolutional neural network on the embedded platform. For this, we first analyzed the time-consuming layers in the inference process of the network, and then we design optimization methods for these layers. Also, we design a memory pool specifically for neural networks. Our experimental results show that our embedded inference framework can run a classification model MobileNet in 80ms and a detection model MobileNet-SSD in 155ms on Firefly-RK3399 development board.

Mobile sensing receives lots of attention since it exploits sensors embedded in smartphones to gather information for quality improvement of daily life. In particular, mobile users (MUs) carrying smartphones act as queriers who request information provided by MUs who act as data collectors, and the data collection and delivery are facilitated by human mobility and ubiquity. However, additional privacy considerations rise in this paradigm, where MUs might not want to reveal their identities when providing some information at a specific position during a particular period. By adopting group signature, where any member of its group can sign a message on behalf of the entire group, the signer of the sensed information cannot be recognized, and thus, anonymity is preserved. The concept of group signature can be realized when introducing mobile edge computing, where base stations are clustered into a region to serve a group of MUs. In such region-based group signature, MUs located in the same region are grouped and thus are indistinguishable. Due to the localized movement feature of human, MUs might not frequently leave the current group and join a new group, which subsequently reduces the infamous rekeying overhead. We implement group signature on Android smartphones and conduct simulation experiments to investigate the performance of the proposed region-based grouping approach from the perspectives of rekeying overhead and delay. The simulation results show that the region-based grouping method is efficient. Moreover, the optimal group number for the minimization of rekeying overhead is suggested.

Mobile sensing techniques have been increasingly deployed in many Internet of Things-based applications because of their cost efficiency, wide coverage, and flexibility. However, these techniques are unreliable in many situations due to noise of different kinds, loss of communication, or insufficient energy. As such, datasets created from mobile sensing scenarios are likely to contain large amount of missing data, which makes further data analysis difficult, inaccurate, or even impossible. We find that the existing estimation models and techniques developed for static sensing do not work well in the mobile sensing scenarios. To address the problem, we propose a spatio-temporal method, which is specifically designed for answering queries in such applications. Experiments on a real-world, incomplete mobile sensing dataset show that the proposed method outperforms the state-of-the-art noticeably in terms of estimation errors. More importantly, the proposed model is tolerant to datasets with extremely high-missing data rates. Training with the proposed model is also efficient, which makes it suitable for deployment on computationally constrained devices and platforms that need to process massive amounts of data in real time.

Today's mobile phone users are faced with large numbers of notifications on social media, ranging from new followers on Twitter and emails to messages received from WhatsApp and Facebook. These digital alerts continuously disrupt activities through instant calls for attention. This paper examines closely the way everyday users interact with notifications and their impact on users' emotion. Fifty users were recruited to download our application NotiMind and use it over a five-week period. Users' phones collected thousands of social and system notifications along with affect data collected via self-reported Positive and Negative Affect Schedule tests three times a day. Results showed a noticeable correlation between positive affective measures and keyboard activities. When large numbers of post and remove notifications occur, a corresponding increase in negative affective measures is detected. Our predictive model has achieved a good accuracy level using three different "in the wild" classifiers (F-measure 74%-78% within-subject model, 72%-76% global model). Our findings show that it is possible to automatically predict when people are experiencing positive, neutral, or negative affective states based on interactions with notifications. We also show how our findings open the door to a wide range of applications in relation to emotion awareness on social and mobile communication.