SEP/BA: Design of a GUI for Definition and Manipulation of Hidden Markov Models
Advisor: Nassir Navab
Supervision by: Nicolas Padoy
, Tobias Blum
Hidden Markov Models (HMMs) are widely used for statistical modeling all kind of time series. They are used for speech recognition, gesture recognition, in bioinformatics and many other domains. Our main interest is in using HMMs to model human activities during surgeries. Here, HMMs are used on different kinds of data that can be obtained during a surgery, e.g. accelerometer data, information about instrument use and signals obtained by computer vision methods. Such a HMM can be used to automatically estimate the current phase of a surgery and provide a context-sensitive user interface. Other applications include monitoring and prediction of the remaining duration of a surgery.
A big challenge when using HMMs is to find an appropriate topology. In this work a HMM editor will be developed, that can be used to manually design and manipulate HMM topologies. Such an editor can also be used to explore the statistical properties, embodied in an HMM.
During this thesis you have to implement a graphical user interface and learn about the basic algorithms for training and using a HMM. You will also get insight into the workflow of surgeries and you also have the possibility to attend real surgeries. Previous knowledge in machine learning and designing user interfaces in C++ are helpful but not mandatory.
You will find related projects on our workflow project page
LR. Rabiner. A tutorial on hidden Markov models and selected applications in speech recognition, Proceedings of the IEEE 77(2)
If you are interested please contact Nicolas Padoy
or Tobias Blum