Semester and Master theses
Mental/Social Recognition - FS12

---

Tweather - using Twitter to generate a weather map [D-ITET/D-INFK, 1]

We often talk about the weather and it's no difference in social networking services like Twitter. In fact, around 0.3% of all Twitter messages contain weather observations. That's around 500'000 messages per day. The aim of this work is to investigate if we use such Twitter messages to obtain a clear picture of the current weather situation and generate a real-time weather map by mining geo-tagged Twitter messages.

This project is part of the EU project SOCIONICAL (www.socionical.eu)

Distribution: 25% Theory, 25% Development, 50% Evaluation
Requirements: Experience in Python, data mining
Main supervisor: Martin Wirz, H97, E-Mail: martin.wirz@ife.ee.ethz.ch, Telephone: 25829
Second supervisor: Sebastian Feese, H67, E-Mail: feese@ife.ee.ethz.ch, Telephone: +41 44 632 30 77
Project Title:
Professor: Prof. Tröster

---

Extracting movement patterns from stair climbing for monitoring of the elderly [D-ITET/D-INFK, 1]

Due to the aging society elderly care is an up-to-date research topic. We focus on assisting elderly people with living as long as possible in their familiar environment using sensor technology. Within this monitoring, a quantitative measure of the patient's health and fitness level is of great interest to doctors and caretakers.
Your task would be to identify different movement patterns to assess the skill of stair climbing for the evaluation of the subject's skill and thus health or fitness level. These will be collected during an experiment with body-worn sensors. Machine learning techniques will be used to extract the most relevant information from the sensor data. Several algorithms should be compared in terms of performance, sensor positioning, and execution time. Measurement hardware will be provided to support your work.


Distribution: 20% Lliterature - 60% Algorithms - 20% Experiment & Evaluation
Requirements: Matlab, machine learning
Main supervisor: Julia Seiter, H97, E-Mail: julia.seiter@ife.ee.ethz.ch, Telephone: 22744
Second supervisor: Christina Strohrmann, H 64, E-Mail: strohrmann@ife.ee.ethz.ch, Telephone: +41 44 632 05 44
Project Title:
Professor: Prof. Tröster

---

Inferring health condition from movement patterns - a user study [D-ITET/D-INFK, 1]

The aging society introduces several difficulties into elderly care, one of them being the increasing number of elderly people to monitor on a regular basis. We work towards applying sensor technology to support doctors and caretakers with automatic monitoring of elderly people in their everyday environment.
During stair climbing, different movement patterns and the stair climbing speed indicate a subject's health and fitness level. You will work towards automatic assessment of the stair climbing task. In a literature review and interviews with clinicians you will identify the most prominent movement patterns. Afterwards, you will collect data covering these patterns using body-worn sensors and develop an algorithm to assess these. A user study will be performed to evaluate your algorithm.


Distribution: 20% Literature - 30% Algorithms - 50% Experiment & Evaluation
Requirements: Matlab, interest in experiments, machine learning beneficial
Main supervisor: Julia Seiter, H97, E-Mail: julia.seiter@ife.ee.ethz.ch, Telephone: 22744
Second supervisor: Christina Strohrmann, H 64, E-Mail: strohrmann@ife.ee.ethz.ch, Telephone: +41 44 632 05 44
Project Title:
Professor: Prof. Tröster

---

How Fast Are You? Wearable Reaction Time Tests in Everyday Life [D-ITET/D-INFK, 1..2]

Did you know that a simple reaction time (RT) test is a valuable measure to detect mild cognitive loss, Alzheimer's disease or Attention Deficit Hyperactive Disorder (ADHD)? However, a main drawback of existing RT tests is that they require the full attention of a test person which prohibits the measurement of cognitive functioning during daily routine tasks. In order to overcome these limitations, in a former semester thesis we have developed a wearable on-body sensor system which measures the reaction time of the user throughout everyday life activities. In our wearable device, we combined the generation of a haptic stimulus (vibration motor on wrist) and gesture recognition (hand movements detected by accelerometers). By computing the time between the vibration stimulus and hand movement response to it, we can measure one's reaction times during everyday life.
The main goal of this thesis is to design and conduct a real life experiment by using wearable reaction time tests in order to measure one's cognitive functioning. In a first step, the existing micro-controller software has to be extended in a way that the generation of vibration stimuli relies on the context of the user, e.g. only during calm activities (measured by the accelerometer) the stimuli should be generated. In a second step, a real life experiment is conducted in which reaction times and error rates are collected from the wearable devices. In addition, subjective performance measures (questionnaire data on perceived concentration and alertness) will be collected. Afterwards, the relation between reaction time and perceived cognitive performance in daily life will be examined.


Distribution: 60% Experiments, 40% Software
Requirements: Interest in experiments, Programming is beneficial
Main supervisor: Burcu Cinaz, H67, E-Mail: burcu.cinaz@ife.ee.ethz.ch, Telephone: 20272
Second supervisor: Bert Arnrich, H93, E-Mail: barnrich@ife.ee.ethz.ch, Telephone: 25280
Project Title:
Professor: Prof. Tröster

---

Gesture-Based Experience Sampling Method: Completing Questionnaires with a Wrist-Worn Device [D-ITET/D-INFK, 1..2]

In many user studies it is important to collect self-assessments on subject?s feelings, thoughts or behaviors at certain times. A common method is the so called Experience Sampling Method (ESM) where the subject is randomly prompted during his daily life to complete a set of questionnaires. Mostly the questionnaires are filled on a paper, computer or mobile phone. The goal of this work is to investigate hand gestures as an alternative input modality to provide basic self-assessments on mood and arousal. Until now, we implemented a wrist-worn wearable device which is able to automatically generate a haptic stimulus and to continuously collect acceleration data. In this work, first the existing micro-controller software of the wearable device has to be adapted in order to generate haptic stimuli pattern whenever the subject should complete a mood questionnaire. Next, gesture recognition methods have to be implemented in order to automatically detect subject?s response, e.g. particular hand gestures such as thumps up, thumps down to express happiness or sadness or hand movements with different velocities to express anger or excitement. After implementing a gesture recognition system, an experiment with different subjects will be done in which the performance of the system will be compared with a classical paper-based or button-based solution.

Distribution: %60 Software & Algorithms, %40 Experiment
Requirements: Programming skills (Matlab or C), Interest in Experiments
Main supervisor: Burcu Cinaz, H67, E-Mail: burcu.cinaz@ife.ee.ethz.ch, Telephone: 20272
Second supervisor: Amir Muaremi, H60.1, E-Mail: muaremi@ife.ee.ethz.ch, Telephone: +41 44 632 05 44
Project Title:
Professor: Prof. Tröster

---

Buzz - using mobile data to pinpoint night life hot spots [D-ITET/D-INFK, 1]

Do you often ask yourself the traditional Friday night question: Where should I go? What's going on? Which is the best place to go? In this work, we would like to use mobile data to pinpoint night life hot spots.

The aim is to develop methods for detecting densely populated areas in a city and visualize this information on a mobile device. GPS-enabled mobile phones should provide the data.

Since not every person is willing to share their location, a full coverage is likely not possible. Hence, a major contribution of your work is to develop methods which work even if only a few people participate. In a second part of your work, you would test your system methods in a real situation and analyze the performance and limitations of your approach.

In a previous work, we have developed a system to simultaneously agregate sensor information of multiple mobile phones. This system should be used for data aquisation and integration of the developed methods.

This project is part of the EU project SOCIONICAL (www.socionical.eu)


Distribution: 60% Evaluation & Experiments, 40% Software
Requirements: Experience in mobile software development
Main supervisor: Martin Wirz, H97, E-Mail: martin.wirz@ife.ee.ethz.ch, Telephone: 25829
Second supervisor: Sebastian Feese, H67, E-Mail: feese@ife.ee.ethz.ch, Telephone: +41 44 632 30 77
Project Title:
Professor: Prof. Tröster

---

Multimodal Conversation Analysis: Linking Body Language and Meaning [D-ITET/D-INFK, 1]

Face to face conversations are at the heart of our social lives. Whenever we interact with others we not only communicate verbally but also nonverbally through facial expressions, gestures, postures and other body movements. It is the goal of this thesis to develop and evaluate methods to automatically use speech (e.g. talking time) and body movement data (head nods, hand gestures, head orientation) to cluster a conversation into typical phases such as for example monolog and dialog. In a second step, you will analyze the available speech transcriptions to cluster speech acts into semantic topics that you will link to sensor data. In your analysis you will evaluate your methods in terms of stability and robustness on a large multi-modal data set is which is already available.

Distribution: 20% Literature, 50% Algorithms, 30% Evaluation
Requirements: Programming in Matlab or R, Machine Learning
Main supervisor: Sebastian Feese, H67, E-Mail: feese@ife.ee.ethz.ch, Telephone: +41 44 632 30 77
Second supervisor: Martin Wirz, H97, E-Mail: martin.wirz@ife.ee.ethz.ch, Telephone: 25829
Project Title:
Professor: Prof. Tröster