Exploration of multi modal datasets towards sleep quality analysis

Over the last years, several datasets on stress and mental health in diverse populations were collected at imec. Collected data types include activity, logs of feelings, sleep quality,  sleep and stress annotations and physiological data from wearables (ECG signals, skin conductance, activity…). Additional to our internal development on algorithms for behaviour modeling, we are searching for novel ways to mine multi-modal datasets and uncover new insights from the data, for this proposal we want to focus on application towards sleep analysis.

In this thesis/internship, sleep quality will be assessed using movement, heart rate, electrodermal activity and skin temperature. This multitude of signals will allow for a unique approach on sleep quality analysis. Especially regarding electrodermal activity, there is a growing interest towards the appearance of so-called ‘EDA-storms’ overnight and how they relate to skin temperature increases. The student will work on multiple existing databases and is expected to apply machine learning and/or statistical modelling techniques to identify interesting patterns.

Specific thesis objectives:

• Study of the state of the art of sleep quality monitoring techniques using wearable sensors
• Develop and test a sleep detection algorithm based on  a public dataset and  imec’s datasets.
• Build a model for sleep quality analysis using  imec’s dataset on healthy persons

Requirements:

• Interest and enthusiasm in data analysis of physiological signals from wearable devices (ECG, skin conductance, activity,...)
• Basic knowledge of statistical  modelling and/or machine learning techniques
• Knowledge of Matlab (R and Python are a plus)
• Eager to learn new analysis techniques