Back to list

Computer-assisted classification of tremors based on acceleration data

Authors Weinmüller, B.
Year 2017
Thesis Type Master's thesis
Topic Sonification
Keywords Music Information Retrieval
Abstract Today tremors are the most common movement disorder and are defined as a rhythmic and unintentional shaking of body part, which is evoked by a repeated tightening of contrarious muscle groups. As the diagnosis is extremely time consuming, it is the aim of this thesis to explore an assisting diagnosis tool to help general practitioners in the first diagnosis. With this assisting tool it should become possible to create categories that will be created exclusively by sensor signals. The sensor data was recorded by acceleration sensors which were placed on the hands of the patients. Data was collected in two groups of positions (rest and posture). The goal is to find a model in which the analysed data would be automatically correctly classified. Different signal analysis methods, like principal component analysis and structure analysis or respectively quantification analysis of recurrence plots, were used to extract features of the data, which then allowed a classification. As a previous data-sonification delivered interesting results, alternative features from music information retrieval were also used in this thesis.
Supervisors Sontacchi, A.