Short Description

Recently, there was increasing interest in analyzing various information gained by smartphones. By detecting variations in the individual user data one can infer actual user behavior – opening the door for mobile health solutions. Therefore, it would be interesting to reliably detect sudden variation of the underlying data properties, i.e., change points.

Change point detection can be formulated as segmenting a time series into partitions with different statistics [1]. Similar work has been extended to online detection of change points [2].

A framework for logging and providing smartphone user data is available.

Your Tasks

• Literature survey: change point detection in multivariate time-series
• Implement and compare the algorithms from literature
• Apply change point detection to smartphone user-data

Your Profile

• Motivation and interest in the topic
• Background in signal processing and/or machine learning
• Experience in Python or Matlab

Additional Information

This thesis is a cooperation with the start-up meemo-tec. Meemo-tec develops a smartphone based health assistive system for bipolar people. The outcome of this thesis will be used to improve the detection of changes in the state of mental health.

For master students a monthly financial compensation will be provided.

References

[1] J.I. Takeuchi and K. Yamanishi: A Unifying Framework for Detecting Outliers and Change Points from Time Series.

[2] R.P. Adams and D.J. MacKay: Bayesian Online Changepoint Detection.