Diaconița, Irina (2018)
Context-aware phone mode adaptation – Approaches to Classify Phone Position and User Activity from Smartphone Sensor Data.
Technische Universität Darmstadt
Ph.D. Thesis, Primary publication
|
Text
2018-04-18_Diaconita_Irina.pdf - Published Version Copyright Information: CC BY-NC-ND 4.0 International - Creative Commons, Attribution NonCommercial, NoDerivs. Download (2MB) | Preview |
Item Type: | Ph.D. Thesis | ||||
---|---|---|---|---|---|
Type of entry: | Primary publication | ||||
Title: | Context-aware phone mode adaptation – Approaches to Classify Phone Position and User Activity from Smartphone Sensor Data | ||||
Language: | English | ||||
Referees: | Steinmetz, Prof. Dr. Ralf ; Mauthe, Prof. Dr. Andreas | ||||
Date: | 26 March 2018 | ||||
Place of Publication: | Darmstadt | ||||
Date of oral examination: | 26 March 2018 | ||||
Abstract: | Over the past few years both the popularity and the computing and sensing capabilities of smartphones increased significantly. This created tremendous opportunities for pervasive and mobile computing. Previous approaches had to rely on users carrying dedicated devices which gained limited acceptance and spread due to the constraints and extra costs involved. Nowadays a great and ever-increasing number of users own and carry smartphones, thus enabling the monitoring of user context and behavior and the provision of accordingly adapted services while not disturbing the user. Despite the plethora of features that smartphones provide, a significant proportion of their usage still belongs to communications appplications [9]. In this context, adequate adaptation of the phone notification mechanisms is necessary in order to avoid disruptions and disturbances to the user and surrounding peers, as well as in order to ensure that notifications are not mistakenly overseen. The survey we carried out indicates that a significant percentage of respondents at least occasionally fail to adapt their phone mode, whereas the majority of the participants report having to change their phone modes several times per day. The goal of this thesis is to provide the users with methods to automatically adapt their phones’ notification mechanism based on two types of contextual information that we deemed as essential: the place where the phone is carried and the users’ current activities. Phone position detection constitutes our main focus regarding context detection, while for user activity classification we adapt existing approaches to our scenarios. Both types of contextual information are inferred only based on hardware sensors embedded in off-the-shelf smartphones in order to make this approach available to as many users as possible. The main challenges include obtaining the highest possible classification accuracy and ensuring that the contextual information is not deprecated, given the great risk of disturbance posed by inaccurate results. Furthermore, we aim to limit the duration of sensor sampling in order to reduce both battery impact and privacy concerns. Our contributions include an audio-based active probing approach which achieves a significant improvement of phone position classification accuracy in comparison to existing state-of-the-art approaches. We subsequently improved this method, which relies on playing audio signals and recording them at the same time, by opportunistically piggybacking phone notification sounds. Thus we could at the same time avoid user disturbance through sound playing, limit the recording duration, and ensure that the classified phone position is actually the current one. For the situations when the phone is on silent mode, we propose an approach that piggybacks vibration motor movements to collect and classify accelerometer readings, achieving accuracies in excess of 90%. Lastly, we provide a proof of concept implementation of a system that awaits incoming phonecalls or notifications, records audio and accelerometer readings, determines the user’s activity and phone position, and immediately adapts the phone mode. |
||||
Alternative Abstract: |
|
||||
URN: | urn:nbn:de:tuda-tuprints-73491 | ||||
Classification DDC: | 000 Generalities, computers, information > 004 Computer science | ||||
Divisions: | 18 Department of Electrical Engineering and Information Technology | ||||
Date Deposited: | 29 May 2018 14:16 | ||||
Last Modified: | 09 Jul 2020 02:04 | ||||
URI: | https://tuprints.ulb.tu-darmstadt.de/id/eprint/7349 | ||||
PPN: | 432152873 | ||||
Export: |
View Item |