Ozlem Incel | Galatasaray University (original) (raw)

Papers by Ozlem Incel

birthdays together. That was funny that we attended a class on "Turkish" folklore dance in "the N... more birthdays together. That was funny that we attended a class on "Turkish" folklore dance in "the Netherlands"! Our picture with the traditional costumes is still hanging on the board in the Pervasive Systems corridor. I'd like to thank Ayşegül for the great memories, for her care and support during this journey and for her help by being my paranymph. Ştefan and Ileana, the other Romanian couple, have also been very good friends to have very nice time together. The parties, gaming gatherings, volleyball sessions wouldn't have been this lively without their presence. Ş tefan helped me a lot to progress in my research studies especially at the very beginning. He is a great Matlab master and taught me how to collect fast and efficient simulation results. I'd like to thank both of them for their warm friendship. Supriyo... We both still complain that we couldn't have collaborated enough, so far, but I believe that we will continue interacting in the future. His enthusiasm, being always ready to help or discuss about anything have made a very lively and positive office environment. His reviews on my introduction chapter helped me how to restructure the rest of the thesis. He was the one who proposed to visit Bhaskar's group when we went to California in 2007. This visit gave me the opportunity to clarify my research visit to Anrg. I'd like to thank Supriyo, accompanied with Anindita and little Samhita, for making the life more livelier. Nirvana, I call her the problem solver. If you ask her a question about anything, about the group, management issues, etc., she always has the answer or she knows who can answer it. I thank her many times for making things easier and for her continuous support and for her positiveness with her warm smile. Other former/current members of the Pervasive Systems group have provided valuable feedback on my studies. Hoping not to forget any of the names;

Physical activity recognition using embedded sensors has enabled many context-aware applications ... more Physical activity recognition using embedded sensors has enabled many context-aware applications in different areas, such as healthcare. Initially, one or more dedicated wearable sensors were used for such applications. However, recently, many researchers started using mobile phones for this purpose, since these ubiquitous devices are equipped with various sensors, ranging from accelerometers to magnetic field sensors. In most of the current studies, sensor data collected for activity recognition are analyzed offline using machine learning tools. However, there is now a trend towards implementing activity recognition systems on these devices in an online manner, since modern mobile phones have become more powerful in terms of available resources, such as CPU, memory and battery. The research on offline activity recognition has been reviewed in several earlier studies in detail. However, work done on online activity recognition is still in its infancy and is yet to be reviewed. In this paper, we review the studies done so far that implement activity recognition systems on mobile phones and use only their on-board sensors. We discuss various aspects of these studies. Moreover, we discuss their limitations and present various recommendations for future research.

IEEE Access

Authenticating a user in the right way is essential to IT systems, where the risks are becoming m... more 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.

Communications in Computer and Information Science

Modern mobile devices are capable of sensing a large variety of changes, ranging from users' moti... more Modern mobile devices are capable of sensing a large variety of changes, ranging from users' motions to environmental conditions. Context-aware applications utilize the sensing capability of these devices for various purposes, such as human activity recognition, health coaching or advertising, etc. Identifying devices and authenticating unique users is another application area where mobile device sensors can be utilized to ensure more intelligent, robust and reliable systems. Traditional systems use cookies, hardware or software fingerprinting to identify a user but due to privacy and security vulnerabilities, none of these methods propose a permanent solution, thus sensor fingerprinting not only identifies devices but also makes it possible to create non-erasable fingerprints. In this work, we focus on distinguishing devices via mobile device sensors. To this end, a large dataset, larger than 25 GB, which consists of accelerometer and gyroscope sensor data from 21 distinct devices is utilized. We employ different classification methods on extracted 40 features based on various time windows from mobile sensors. Namely, we use random forest, gradient boosting machine, and generalized linear model classifiers. In conclusion, we obtain the highest accuracy as 97% from various experiments in identifying 21 devices using gradient boosting machine on the data from accelerometer and gyroscope sensors.

2017 10th International Conference on Electrical and Electronics Engineering (ELECO), 2017

Time series forecasting is currently used in various areas. Energy management is also one of the ... more Time series forecasting is currently used in various areas. Energy management is also one of the most prevalent application areas. As a matter of fact, energy suppliers and managers have to face with the energy mix problem. Electricity can be produced from fossil fuels, from nuclear energy, from bio-fuels or from renewable energy resources. Concerning electricity generation system based on solar irradiation, it is very important to know precisely the amount of electricity available for the different sources and at different horizons: minutes, hours, and days. Depending on the horizon, two main classes of methods can be used to forecast the solar irradiation: statistical time series forecasting methods for short to midterm horizons and numerical weather prediction methods for medium to long-term horizons. On this paper we focus only on statistical time series forecasting methods. The aim of this study is to assess if deep learning can be suitable and competitive on the solar irradiat...

Augmented reality (AR) is a type of virtual reality aiming to duplicate real world’s environment ... more Augmented reality (AR) is a type of virtual reality aiming to duplicate real world’s environment on a computer’s video feed. The mobile application, which is built for this project (called SARAS), enables annotating real world point of interests (POIs) that are located near mobile user. In this paper, we aim at introducing a robust and simple algorithm for placing labels in an augmented reality system. The system places labels of the POIs on the mobile device screen whose GPS coordinates are given. The proposed algorithm is compared to an existing one in terms of energy consumption and accuracy. The results show that the proposed algorithm gives better results in energy consumption and accuracy while standing still, and acceptably accurate results when driving. The technique provides benefits to AR browsers with its open access algorithm. Going forward, the algorithm will be improved to more rapidly react to position changes while driving.

With the help of the sensors available on smart phones and smart watches, inference on user conte... more With the help of the sensors available on smart phones and smart watches, inference on user context, particularly the activity, can be inferred. Raw data collected from the sensors enables the classification of human activities with machine learning algorithms. In studies focusing on mile activity recognition, usually the motion sensors, such as accelerometer and gyroscope are used. In this paper, our aim is to analyze the performance of activity classification when these sensors are used separately or in combination. By using a dataset which is collected from fifteen participants including six different activities, we extract various features from raw data and afterwards supervised machine learning algorithms are used to train and validate the results. Five different classifiers and different validation methods are used for performance analysis.

TURKISH JOURNAL OF ELECTRICAL ENGINEERING & COMPUTER SCIENCES

Motion sensors available on smart phones make it possible to recognize human activities. Accelero... more Motion sensors available on smart phones make it possible to recognize human activities. Accelerometer, gyroscope, magnetometer, and their various combinations are used to classify, particularly, locomotion activities, ranging from walking to biking. In most of the studies, the focus is on the collection of data and on the analysis of the impact of different parameters on the recognition performance. The parameter space includes the types of sensors used, features, classification algorithms, and position/orientation of the mobile device. In most of the studies, the impact of some of these parameters is partially analyzed; however, in this work, we investigate the parameter space in detail with a global focus. Particularly, we investigate the impact of using different feature-sets, the impact of using different sensors individually and in combination, the impact of different classifiers, and the impact of phone position. Using an ANOVA analysis, we investigate the importance of various parameters on the recognition performance. We show that these parameters are ranked according to their impact on the recognition performance in the following order: sensor, position, classifier, feature. We believe that such an analysis is important since we can statistically show how much a parameter is affecting the recognition performance. Our observations can be used in future studies by only focusing on the important parameters. We present our findings as a discussion to guide the further studies in this domain.

Procedia Computer Science

Abstract Mobile banking applications are one of the most sensitive apps for secure authentication... more Abstract Mobile banking applications are one of the most sensitive apps for secure authentication. Recently, use of continuous authentication using behavioral metrics are proposed where interactions of users with the mobile devices are tracked by collecting sensor data and touch-screen data. However, this may bring extra overhead in terms of resource consumption considering battery limited devices. In this work, augmentation of a mobile banking application with continuous authentication using behavioral biometrics is presented and its performance is analyzed in terms of resource usage. The banking application is running on Android phones and the presented architecture uses the data generated by the accelerometer, the gyroscope and the magnetometer sensors and the touchscreen usage. We examined the power consumption, CPU usage, I/O usage and the impacts of the sampling rate of each sensor. We also proposed possible performance improvements as future work.

TURKISH JOURNAL OF ELECTRICAL ENGINEERING & COMPUTER SCIENCES

Energy management is an emerging problem nowadays and utilization of renewable energy sources is ... more Energy management is an emerging problem nowadays and utilization of renewable energy sources is an efficient solution. Solar radiation is an important source for electricity generation. For effective utilization, it is important to know precisely the amount from different sources and at different horizons: minutes, hours, and days. Depending on the horizon, two main classes of methods can be used to forecast the solar radiation: statistical time series forecasting methods for short to midterm horizons and numerical weather prediction methods for medium-to long-term horizons. Although statistical time series forecasting methods are utilized in the literature, there are a limited number of studies that utilize deep artificial neural networks. In this study, we focus on statistical time series forecasting methods for short-term horizons (1 h). The aim of this study is to discover the effect of using multivariate data on solar radiation forecasting using a deep learning approach. In this context, we propose a multivariate forecast model that uses a combination of different meteorological variables, such as temperature, humidity, and nebulosity. In the proposed model, recurrent neural network (RNN) variation, namely a long short-term memory (LSTM) unit is used. With an experimental approach, the effect of each meteorological variable is investigated. By hyperparameter tuning, optimal parameters are found in order to construct the best models that fit the global solar radiation data. We compared the results with those of previous studies and we found that the multivariate approach performed better than the previous univariate models did. In further experiments, the effect of combining the most effective parameters was investigated and, as a result, we observed that temperature and nebulosity are the most effective parameters for predicting future solar radiance.

Procedia Computer Science

Abstract In this paper, we present the ARService framework which is a crowd-sourced mobile sensin... more Abstract In this paper, we present the ARService framework which is a crowd-sourced mobile sensing system with an online activity recognition module running on a smartphone. The system consists of a mobile application and a server part. The application logs data from sensors, particularly motion sensors, available on smartphones, data about the phone state, such as battery level, location information, as well as data from the wireless interfaces, such as the nearby access points. Besides being a data logger, the application also recognizes user activities, such as walking, sitting, using accelerometer in an online manner. On the server side, data is stored for further analysis and also visualized. ARService was continuosly used by 15 participants for a duration of one month in a data collection campaign. Besides the details of the framework, we present the online activity recognition performance and show that, up to 89% accuracy is achieved in recognizing the activities of the participants in an online manner.

Journal of Ambient Intelligence and Humanized Computing

Semantic place prediction problem is the process of giving semantic names to locations. While the... more Semantic place prediction problem is the process of giving semantic names to locations. While the localization problem is about predicting the exact position, i.e. the coordinates, of a place, the aim here is to semantically characterize the location, such as home, school, restaurant. In order to solve the problem, phone usage patterns of crowds and the performed activities at different places can be utilized. In this study, we aim to semantically classify places visited by smart phone users utilizing the data collected from sensors and wireless interfaces available on the phones as well as phone usage patterns, such as battery level, and time-related information, with machine learning algorithms. For this purpose, we collected data from 15 participants at Galatasaray University for a duration of 1 month, in April 2016, and two of the users continued to collect 1 more month of data, which makes it 17 participants in total. We extract various set of features from the collected data and analyse the efficiency of features with different classification algorithms such as, decision tree, random forest, k-nearest neighbour, naive Bayes and multi-layer perceptron. We observe that, by fusing features extracted from different sources of data, better success rates are achieved. Moreover, we explore the relationship between places and activities, which was not explored in previous studies, and show that activities are important source of information for characterizing the places. Additionally, we observe that, while a generalized classifier performs reasonably well, using person-specific data and classification can help to improve the success rate.

International Journal of Communication Systems, 2016

Procedia Computer Science, 2016

In this work, design of an Android-based augmented reality application is presented and particula... more In this work, design of an Android-based augmented reality application is presented and particularly its performance is analyzed in terms of resource usage in comparison to similar applications. The application displays merchant, branch information of one of the Turkish banks, as well as related sales campaigns of the merchants on the screen that are within the proximity of the user's location. The developed application uses GPS, compass, gyroscope, accelerometer sensors and it utilizes an accurate tagging algorithm. We examine the resource and battery consumption of the application. Accordingly, we propose methods for improving the resource usage. The proposed improvements reduce the resource consumptions up to 35% and the application performs considerably well compared to the state of the art commercial applications. We believe that the suggested improvements can be useful for other sensor-based mobile augmented reality applications.

Computer Networks the International Journal of Computer and Telecommunications Networking, Sep 1, 2011

Sensors, 2016

The position of on-body motion sensors plays an important role in human activity recognition. Mos... more The position of on-body motion sensors plays an important role in human activity recognition. Most often, mobile phone sensors at the trouser pocket or an equivalent position are used for this purpose. However, this position is not suitable for recognizing activities that involve hand gestures, such as smoking, eating, drinking coffee and giving a talk. To recognize such activities, wrist-worn motion sensors are used. However, these two positions are mainly used in isolation. To use richer context information, we evaluate three motion sensors (accelerometer, gyroscope and linear acceleration sensor) at both wrist and pocket positions. Using three classifiers, we show that the combination of these two positions outperforms the wrist position alone, mainly at smaller segmentation windows. Another problem is that less-repetitive activities, such as smoking, eating, giving a talk and drinking coffee, cannot be recognized easily at smaller segmentation windows unlike repetitive activities, like walking, jogging and biking. For this purpose, we evaluate the effect of seven window sizes (2-30 s) on thirteen activities and show how increasing window size affects these various activities in different ways. We also propose various optimizations to further improve the recognition of these activities. For reproducibility, we make our dataset publicly available.

Sensors, 2015

Phone placement, i.e., where the phone is carried/stored, is an important source of information f... more Phone placement, i.e., where the phone is carried/stored, is an important source of information for context-aware applications. Extracting information from the integrated smart phone sensors, such as motion, light and proximity, is a common technique for phone placement detection. In this paper, the efficiency of an accelerometer-only solution is explored, and it is investigated whether the phone position can be detected with high accuracy by analyzing the movement, orientation and rotation changes. The impact of these changes on the performance is analyzed individually and both in combination to explore which features are more efficient, whether they should be fused and, if yes, how they should be fused. Using three different datasets, collected from 35 people from eight different positions, the performance of different classification algorithms is explored. It is shown that while utilizing only motion information can achieve accuracies around 70%, this ratio increases up to 85% by utilizing information also from orientation and rotation changes. The performance of an accelerometer-only solution is compared to solutions where linear acceleration, gyroscope and magnetic field sensors are used, and it is shown that the accelerometer-only solution performs as well as utilizing other sensing information. Hence, it is not necessary to use extra sensing information where battery power consumption may increase. Additionally, I explore the impact of the performed activities on position recognition and show that the accelerometer-only solution can achieve 80% recognition accuracy with stationary activities where movement data are very limited. Finally, other phone placement problems, such as in-pocket and on-body detections, are also investigated, and higher accuracies, ranging from 88% to 93%, are reported, with an accelerometer-only solution.

The ability of properly covering the terrain under investigation and collecting measurements from... more The ability of properly covering the terrain under investigation and collecting measurements from re-dundantly sensed portions of the terrain are two important objectives for monitoring applications using wireless sensor networks. Here, a new architectural view of information retrieval for XML compliant environmental monitoring ap-plications is introduced. The sensor network is constructed to satisfy the requirements of monitoring applications and maintained as long as the application has queries to be run on the sensor nodes. Mobile clients of the architecture (drivers), such as human beings or autonomous robots, navigate within the sensing environment and build up sensor node trees in order to effectively disseminate queries and collect the results. In the paper, viable methods are proposed for query service binding, query driven sensor network topology construction and end-to-end event delivery on dynamically maintained return paths. Simulation re-sults considering the reliabilit...

This paper introduces a unifying architectural view of information retrieval in environmental mon... more This paper introduces a unifying architectural view of information retrieval in environmental monitoring applications. The architecture incorporates micro sensor nodes and mobile cluster heads into the larger context of a multi-tiered service aware ad hoc backbone network. Mobile clients, such as livings or au- tonomous robots, navigate within the area of interest, query adaptively formed sensor trees and return results to monitoring application via mobile backbone. In the paper, algorithmic solutions for service binding, query driven sensor network topology construction and end-to-end event delivery on dynamically maintained return paths are discussed. Comprehensive experimental results on the backbone topology behavior, sensing area coverage and long term reliability analysis are also given.

birthdays together. That was funny that we attended a class on "Turkish" folklore dance in "the N... more birthdays together. That was funny that we attended a class on "Turkish" folklore dance in "the Netherlands"! Our picture with the traditional costumes is still hanging on the board in the Pervasive Systems corridor. I'd like to thank Ayşegül for the great memories, for her care and support during this journey and for her help by being my paranymph. Ştefan and Ileana, the other Romanian couple, have also been very good friends to have very nice time together. The parties, gaming gatherings, volleyball sessions wouldn't have been this lively without their presence. Ş tefan helped me a lot to progress in my research studies especially at the very beginning. He is a great Matlab master and taught me how to collect fast and efficient simulation results. I'd like to thank both of them for their warm friendship. Supriyo... We both still complain that we couldn't have collaborated enough, so far, but I believe that we will continue interacting in the future. His enthusiasm, being always ready to help or discuss about anything have made a very lively and positive office environment. His reviews on my introduction chapter helped me how to restructure the rest of the thesis. He was the one who proposed to visit Bhaskar's group when we went to California in 2007. This visit gave me the opportunity to clarify my research visit to Anrg. I'd like to thank Supriyo, accompanied with Anindita and little Samhita, for making the life more livelier. Nirvana, I call her the problem solver. If you ask her a question about anything, about the group, management issues, etc., she always has the answer or she knows who can answer it. I thank her many times for making things easier and for her continuous support and for her positiveness with her warm smile. Other former/current members of the Pervasive Systems group have provided valuable feedback on my studies. Hoping not to forget any of the names;

Physical activity recognition using embedded sensors has enabled many context-aware applications ... more Physical activity recognition using embedded sensors has enabled many context-aware applications in different areas, such as healthcare. Initially, one or more dedicated wearable sensors were used for such applications. However, recently, many researchers started using mobile phones for this purpose, since these ubiquitous devices are equipped with various sensors, ranging from accelerometers to magnetic field sensors. In most of the current studies, sensor data collected for activity recognition are analyzed offline using machine learning tools. However, there is now a trend towards implementing activity recognition systems on these devices in an online manner, since modern mobile phones have become more powerful in terms of available resources, such as CPU, memory and battery. The research on offline activity recognition has been reviewed in several earlier studies in detail. However, work done on online activity recognition is still in its infancy and is yet to be reviewed. In this paper, we review the studies done so far that implement activity recognition systems on mobile phones and use only their on-board sensors. We discuss various aspects of these studies. Moreover, we discuss their limitations and present various recommendations for future research.

IEEE Access

Authenticating a user in the right way is essential to IT systems, where the risks are becoming m... more 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.

Communications in Computer and Information Science

Modern mobile devices are capable of sensing a large variety of changes, ranging from users' moti... more Modern mobile devices are capable of sensing a large variety of changes, ranging from users' motions to environmental conditions. Context-aware applications utilize the sensing capability of these devices for various purposes, such as human activity recognition, health coaching or advertising, etc. Identifying devices and authenticating unique users is another application area where mobile device sensors can be utilized to ensure more intelligent, robust and reliable systems. Traditional systems use cookies, hardware or software fingerprinting to identify a user but due to privacy and security vulnerabilities, none of these methods propose a permanent solution, thus sensor fingerprinting not only identifies devices but also makes it possible to create non-erasable fingerprints. In this work, we focus on distinguishing devices via mobile device sensors. To this end, a large dataset, larger than 25 GB, which consists of accelerometer and gyroscope sensor data from 21 distinct devices is utilized. We employ different classification methods on extracted 40 features based on various time windows from mobile sensors. Namely, we use random forest, gradient boosting machine, and generalized linear model classifiers. In conclusion, we obtain the highest accuracy as 97% from various experiments in identifying 21 devices using gradient boosting machine on the data from accelerometer and gyroscope sensors.

2017 10th International Conference on Electrical and Electronics Engineering (ELECO), 2017

Time series forecasting is currently used in various areas. Energy management is also one of the ... more Time series forecasting is currently used in various areas. Energy management is also one of the most prevalent application areas. As a matter of fact, energy suppliers and managers have to face with the energy mix problem. Electricity can be produced from fossil fuels, from nuclear energy, from bio-fuels or from renewable energy resources. Concerning electricity generation system based on solar irradiation, it is very important to know precisely the amount of electricity available for the different sources and at different horizons: minutes, hours, and days. Depending on the horizon, two main classes of methods can be used to forecast the solar irradiation: statistical time series forecasting methods for short to midterm horizons and numerical weather prediction methods for medium to long-term horizons. On this paper we focus only on statistical time series forecasting methods. The aim of this study is to assess if deep learning can be suitable and competitive on the solar irradiat...

Augmented reality (AR) is a type of virtual reality aiming to duplicate real world’s environment ... more Augmented reality (AR) is a type of virtual reality aiming to duplicate real world’s environment on a computer’s video feed. The mobile application, which is built for this project (called SARAS), enables annotating real world point of interests (POIs) that are located near mobile user. In this paper, we aim at introducing a robust and simple algorithm for placing labels in an augmented reality system. The system places labels of the POIs on the mobile device screen whose GPS coordinates are given. The proposed algorithm is compared to an existing one in terms of energy consumption and accuracy. The results show that the proposed algorithm gives better results in energy consumption and accuracy while standing still, and acceptably accurate results when driving. The technique provides benefits to AR browsers with its open access algorithm. Going forward, the algorithm will be improved to more rapidly react to position changes while driving.

With the help of the sensors available on smart phones and smart watches, inference on user conte... more With the help of the sensors available on smart phones and smart watches, inference on user context, particularly the activity, can be inferred. Raw data collected from the sensors enables the classification of human activities with machine learning algorithms. In studies focusing on mile activity recognition, usually the motion sensors, such as accelerometer and gyroscope are used. In this paper, our aim is to analyze the performance of activity classification when these sensors are used separately or in combination. By using a dataset which is collected from fifteen participants including six different activities, we extract various features from raw data and afterwards supervised machine learning algorithms are used to train and validate the results. Five different classifiers and different validation methods are used for performance analysis.

TURKISH JOURNAL OF ELECTRICAL ENGINEERING & COMPUTER SCIENCES

Motion sensors available on smart phones make it possible to recognize human activities. Accelero... more Motion sensors available on smart phones make it possible to recognize human activities. Accelerometer, gyroscope, magnetometer, and their various combinations are used to classify, particularly, locomotion activities, ranging from walking to biking. In most of the studies, the focus is on the collection of data and on the analysis of the impact of different parameters on the recognition performance. The parameter space includes the types of sensors used, features, classification algorithms, and position/orientation of the mobile device. In most of the studies, the impact of some of these parameters is partially analyzed; however, in this work, we investigate the parameter space in detail with a global focus. Particularly, we investigate the impact of using different feature-sets, the impact of using different sensors individually and in combination, the impact of different classifiers, and the impact of phone position. Using an ANOVA analysis, we investigate the importance of various parameters on the recognition performance. We show that these parameters are ranked according to their impact on the recognition performance in the following order: sensor, position, classifier, feature. We believe that such an analysis is important since we can statistically show how much a parameter is affecting the recognition performance. Our observations can be used in future studies by only focusing on the important parameters. We present our findings as a discussion to guide the further studies in this domain.

Procedia Computer Science

Abstract Mobile banking applications are one of the most sensitive apps for secure authentication... more Abstract Mobile banking applications are one of the most sensitive apps for secure authentication. Recently, use of continuous authentication using behavioral metrics are proposed where interactions of users with the mobile devices are tracked by collecting sensor data and touch-screen data. However, this may bring extra overhead in terms of resource consumption considering battery limited devices. In this work, augmentation of a mobile banking application with continuous authentication using behavioral biometrics is presented and its performance is analyzed in terms of resource usage. The banking application is running on Android phones and the presented architecture uses the data generated by the accelerometer, the gyroscope and the magnetometer sensors and the touchscreen usage. We examined the power consumption, CPU usage, I/O usage and the impacts of the sampling rate of each sensor. We also proposed possible performance improvements as future work.

TURKISH JOURNAL OF ELECTRICAL ENGINEERING & COMPUTER SCIENCES

Energy management is an emerging problem nowadays and utilization of renewable energy sources is ... more Energy management is an emerging problem nowadays and utilization of renewable energy sources is an efficient solution. Solar radiation is an important source for electricity generation. For effective utilization, it is important to know precisely the amount from different sources and at different horizons: minutes, hours, and days. Depending on the horizon, two main classes of methods can be used to forecast the solar radiation: statistical time series forecasting methods for short to midterm horizons and numerical weather prediction methods for medium-to long-term horizons. Although statistical time series forecasting methods are utilized in the literature, there are a limited number of studies that utilize deep artificial neural networks. In this study, we focus on statistical time series forecasting methods for short-term horizons (1 h). The aim of this study is to discover the effect of using multivariate data on solar radiation forecasting using a deep learning approach. In this context, we propose a multivariate forecast model that uses a combination of different meteorological variables, such as temperature, humidity, and nebulosity. In the proposed model, recurrent neural network (RNN) variation, namely a long short-term memory (LSTM) unit is used. With an experimental approach, the effect of each meteorological variable is investigated. By hyperparameter tuning, optimal parameters are found in order to construct the best models that fit the global solar radiation data. We compared the results with those of previous studies and we found that the multivariate approach performed better than the previous univariate models did. In further experiments, the effect of combining the most effective parameters was investigated and, as a result, we observed that temperature and nebulosity are the most effective parameters for predicting future solar radiance.

Procedia Computer Science

Abstract In this paper, we present the ARService framework which is a crowd-sourced mobile sensin... more Abstract In this paper, we present the ARService framework which is a crowd-sourced mobile sensing system with an online activity recognition module running on a smartphone. The system consists of a mobile application and a server part. The application logs data from sensors, particularly motion sensors, available on smartphones, data about the phone state, such as battery level, location information, as well as data from the wireless interfaces, such as the nearby access points. Besides being a data logger, the application also recognizes user activities, such as walking, sitting, using accelerometer in an online manner. On the server side, data is stored for further analysis and also visualized. ARService was continuosly used by 15 participants for a duration of one month in a data collection campaign. Besides the details of the framework, we present the online activity recognition performance and show that, up to 89% accuracy is achieved in recognizing the activities of the participants in an online manner.

Journal of Ambient Intelligence and Humanized Computing

Semantic place prediction problem is the process of giving semantic names to locations. While the... more Semantic place prediction problem is the process of giving semantic names to locations. While the localization problem is about predicting the exact position, i.e. the coordinates, of a place, the aim here is to semantically characterize the location, such as home, school, restaurant. In order to solve the problem, phone usage patterns of crowds and the performed activities at different places can be utilized. In this study, we aim to semantically classify places visited by smart phone users utilizing the data collected from sensors and wireless interfaces available on the phones as well as phone usage patterns, such as battery level, and time-related information, with machine learning algorithms. For this purpose, we collected data from 15 participants at Galatasaray University for a duration of 1 month, in April 2016, and two of the users continued to collect 1 more month of data, which makes it 17 participants in total. We extract various set of features from the collected data and analyse the efficiency of features with different classification algorithms such as, decision tree, random forest, k-nearest neighbour, naive Bayes and multi-layer perceptron. We observe that, by fusing features extracted from different sources of data, better success rates are achieved. Moreover, we explore the relationship between places and activities, which was not explored in previous studies, and show that activities are important source of information for characterizing the places. Additionally, we observe that, while a generalized classifier performs reasonably well, using person-specific data and classification can help to improve the success rate.

International Journal of Communication Systems, 2016

Procedia Computer Science, 2016

In this work, design of an Android-based augmented reality application is presented and particula... more In this work, design of an Android-based augmented reality application is presented and particularly its performance is analyzed in terms of resource usage in comparison to similar applications. The application displays merchant, branch information of one of the Turkish banks, as well as related sales campaigns of the merchants on the screen that are within the proximity of the user's location. The developed application uses GPS, compass, gyroscope, accelerometer sensors and it utilizes an accurate tagging algorithm. We examine the resource and battery consumption of the application. Accordingly, we propose methods for improving the resource usage. The proposed improvements reduce the resource consumptions up to 35% and the application performs considerably well compared to the state of the art commercial applications. We believe that the suggested improvements can be useful for other sensor-based mobile augmented reality applications.

Computer Networks the International Journal of Computer and Telecommunications Networking, Sep 1, 2011

Sensors, 2016

The position of on-body motion sensors plays an important role in human activity recognition. Mos... more The position of on-body motion sensors plays an important role in human activity recognition. Most often, mobile phone sensors at the trouser pocket or an equivalent position are used for this purpose. However, this position is not suitable for recognizing activities that involve hand gestures, such as smoking, eating, drinking coffee and giving a talk. To recognize such activities, wrist-worn motion sensors are used. However, these two positions are mainly used in isolation. To use richer context information, we evaluate three motion sensors (accelerometer, gyroscope and linear acceleration sensor) at both wrist and pocket positions. Using three classifiers, we show that the combination of these two positions outperforms the wrist position alone, mainly at smaller segmentation windows. Another problem is that less-repetitive activities, such as smoking, eating, giving a talk and drinking coffee, cannot be recognized easily at smaller segmentation windows unlike repetitive activities, like walking, jogging and biking. For this purpose, we evaluate the effect of seven window sizes (2-30 s) on thirteen activities and show how increasing window size affects these various activities in different ways. We also propose various optimizations to further improve the recognition of these activities. For reproducibility, we make our dataset publicly available.

Sensors, 2015

Phone placement, i.e., where the phone is carried/stored, is an important source of information f... more Phone placement, i.e., where the phone is carried/stored, is an important source of information for context-aware applications. Extracting information from the integrated smart phone sensors, such as motion, light and proximity, is a common technique for phone placement detection. In this paper, the efficiency of an accelerometer-only solution is explored, and it is investigated whether the phone position can be detected with high accuracy by analyzing the movement, orientation and rotation changes. The impact of these changes on the performance is analyzed individually and both in combination to explore which features are more efficient, whether they should be fused and, if yes, how they should be fused. Using three different datasets, collected from 35 people from eight different positions, the performance of different classification algorithms is explored. It is shown that while utilizing only motion information can achieve accuracies around 70%, this ratio increases up to 85% by utilizing information also from orientation and rotation changes. The performance of an accelerometer-only solution is compared to solutions where linear acceleration, gyroscope and magnetic field sensors are used, and it is shown that the accelerometer-only solution performs as well as utilizing other sensing information. Hence, it is not necessary to use extra sensing information where battery power consumption may increase. Additionally, I explore the impact of the performed activities on position recognition and show that the accelerometer-only solution can achieve 80% recognition accuracy with stationary activities where movement data are very limited. Finally, other phone placement problems, such as in-pocket and on-body detections, are also investigated, and higher accuracies, ranging from 88% to 93%, are reported, with an accelerometer-only solution.

The ability of properly covering the terrain under investigation and collecting measurements from... more The ability of properly covering the terrain under investigation and collecting measurements from re-dundantly sensed portions of the terrain are two important objectives for monitoring applications using wireless sensor networks. Here, a new architectural view of information retrieval for XML compliant environmental monitoring ap-plications is introduced. The sensor network is constructed to satisfy the requirements of monitoring applications and maintained as long as the application has queries to be run on the sensor nodes. Mobile clients of the architecture (drivers), such as human beings or autonomous robots, navigate within the sensing environment and build up sensor node trees in order to effectively disseminate queries and collect the results. In the paper, viable methods are proposed for query service binding, query driven sensor network topology construction and end-to-end event delivery on dynamically maintained return paths. Simulation re-sults considering the reliabilit...

This paper introduces a unifying architectural view of information retrieval in environmental mon... more This paper introduces a unifying architectural view of information retrieval in environmental monitoring applications. The architecture incorporates micro sensor nodes and mobile cluster heads into the larger context of a multi-tiered service aware ad hoc backbone network. Mobile clients, such as livings or au- tonomous robots, navigate within the area of interest, query adaptively formed sensor trees and return results to monitoring application via mobile backbone. In the paper, algorithmic solutions for service binding, query driven sensor network topology construction and end-to-end event delivery on dynamically maintained return paths are discussed. Comprehensive experimental results on the backbone topology behavior, sensing area coverage and long term reliability analysis are also given.