My open source project is waiting for your contributions :

https://github.com/yyenigun/mobile-sensing-web

This is an academic web application of Galatasaray University Activity Recognition project for monitoring mobile phone’s sensor data. (used technologies: Java, Spring MVC, Javascript, Bootstrap, Gradle..etc.)

Abstract

Human activity recognition using wireless sensors and crowdsourced sensing are both emerging topics in the domain of pervasive computing. Activity recognition involves the use of different sensing technologies to automatically collect and classify user activities for different application domains, ranging from medical applications to home monitoring. Whereas crowdsourced sensing aims to collect environmental or personal data especially using the mobile devices carried by the people. In this sense, smart phones provide a unique platform both for human activity recognition and crowdsourced sensing applications with the integrated rich set of sensors, such as GPS, accelerometer, microphones, their ubiquity, ease of use and wireless communication capabilities with various interfaces, hence the ability to transfer sensing data to backend servers. In the literature, example crowdsourced sensing applications generally aim to collect geo and time-tagged environmental data, such as noise or air pollution levels in the city. Research on human activity recognition using the sensors available on the phones mostly focuses on the development of applications for the end-user, such as monitoring fitness level of the phone user or the health status. In fact, recognition of the activities of the crowds and communities rather than the activities of individual can enable a new set of application domains in the fields of urban planning, urban transportation, targeted advertising. Hence, it will be an interesting topic to merge these emerging domains of human activity recognition and crowdsourced sensing on the ubiquitous smart phones. One of the objectives of this project is to design and develop an activity based crowdsourced sensing platform where the activities of the individuals related to movement, such as walking, running, as well as their transportation modes, biking, travelling with a vehicle, will be recognized with the sensors on the phones, such as accelerometer, GPS, and the findings will be transmitted to the backend servers where the behaviors of the crowds will be analyzed. This sensing platform will differ from the previous examples in the literature in the sense that the activities will be recognized on the phone rather than offline processing of raw sensor data on the backend servers. Additionally, the platform will support two modes of operation: in the online mode, collected data on the phone is uploaded to a backend server periodically, such as in every 10 minutes to capture the activity dynamics in real-time, whereas in the offline mode, data is uploaded once or twice a day for offline analysis. Additionally, the second objective of the project is to create a large-scale dataset which will be shared with other researchers working in the domain and hence will constitute as a benchmark platform for the comparison of different studies in the field. This dataset will be different from the existing datasets of mobile phone sensing in the sense that the sensors contributing to the activity recognition will be utilized and the data will be collected with newer generations of smartphones which are equipped with a richer set of sensors and provide better capabilities in terms of processing and storage. To achieve these objectives, we have identified the related research challenges and problems which will constitute the topics of the project and which will be addressed following an applied research methodology. The work plan will be divided into four work packages. Particularly, during the project we will first investigate the problems related to the activity recognition process. Selection of machine learning algorithms and the application of those to the mobile phone platform, efficient feature selection techniques for better activity recognition, person and position independent activity recognition will constitute the research problems that will be tackled in this phase. Next, we will investigate the tradeoffs in activity recognition accuracy and the sensor sampling frequency as well as the data transmission frequency and data sizes using optimization algorithms. After solving these problems and developing the required software for the platform, we will proceed with the data collection process. The data collection process will start with the identification of voluntary participants and it will continue for 9 months with at least 20 participants. As the final main research topic of the project, we will investigate the issues related to data processing. The data will be processed with methods such as clustering, and topic models for the identification of places with the activities (semantic tagging) and for the discovery of routine activity patterns for the places. The platform will be tested with a case study where the activities in the university campus will be targeted. The findings of these research problems will be disseminated with research publications, with the objective of 3 conference and 2 journal papers for each year. Additionally, during the project, two meetings will be organized with the objective of sharing the results of the project with other academicians, as well as the industry partners and the municipalities who are identified as the target interest groups who can benefit from the findings of these projects. Researchers who are well-known in the field will be invited to these meetings for sharing their visions on the related problems.