227-0197-00L Wearable Systems I
|Semester||Autumn Semester 2016|
|Lecturers||G. Tröster, U. Blanke|
|Periodicity||yearly recurring course|
|Language of instruction||English|
|Abstract||Context recognition in mobile communication systems like mobile phone, smart watches and wearable computer will be studied using advanced methods from sensor data fusion, pattern recognition, statistics, data mining and machine learning.|
Context comprises the behavior of individuals and of groups, their activites as well as the local and social environment.
|Objective||Using internal sensors and sensors in our environment including data from the wristwatch, bracelet or internet (crowd sourcing), our 'smart phone' detects our context continuously, e.g. where we are, what we are doing, with whom we are together, what is our constitution, what are our needs. Based on this information our 'smart phone' offers us the appropriate services like a personal assistant.Context comprises user's behavior, his activities, his local and social environment.|
In the data path from the sensor level to signal segmentation to the classification of the context, advanced methods of signal processing, pattern recognition and machine learning will be applied. Sensor data generated by crowdsouring methods are integrated. The validation using MATLAB is followed by implementation and testing on a smart phone.
Context recognition as the crucial function of mobile systems is the main focus of the course. Using MatLab the participants implement and verify the discussed methods also using a smart phone.
|Content||Using internal sensors and sensors in our environment including data from the wristwatch, bracelet or internet (crowd sourcing), our 'smart phone' detects our context continuously, e.g. where we are, what we are doing, with whom we are together, what is our constitution, what are our needs. Based on this information our 'smart phone' offers us the appropriate services like a personal assistant. Context recognition - what is the situation of the user, his activity, his environment, how is he doing, what are his needs - as the central functionality of mobile systems constitutes the focus of the course.|
The main topics of the course include
Sensor nets, sensor signal processing, data fusion, time series (segmentation, similariy measures), supervised learning (Bayes Decision Theory, Decision Trees, Random Forest, kNN-Methods, Support Vector Machine, Adaboost, Deep Learning), clustering (k-means, dbscan, topic models), Recommender Systems, Collaborative Filtering, Crowdsourcing.
The exercises show concrete design problems like motion and gesture recognition using distributed sensors, detection of activity patterns and identification of the local environment.
Presentations of the PhD students and the visit at the Wearable Computing Lab introduce in current research topics and international research projects.
Language: german/english (depending on the participants)
|Lecture notes||Lecture notes for all lessons, assignments and solutions. |
|Literature||Literature will be announced during the lessons.|
|Prerequisites / Notice||No special prerequisites|