Oregon State University



Event Details

MS Final Examination – Yonglei Zheng

Friday, August 17, 2012 2:00 PM - 4:00 PM

Predicting Activity Type from Accelerometer Data
The study of the physical activity is important in improving people's health as it can help people understand the relationship between physical activity and health. Accelerometers, due to its small size, low cost, convenience and its ability to provide objective information about the frequency, intensity, and duration of physical activity, has become the method of choice in measuring physical activity. The machine learning algorithms, especially the Artificial Neural Network, based on the featurized representation of the accelerometers data, have become the most widely used approaches in physical activity classification. To improve the classification accuracy, this thesis first explored the impact of the choice of data (raw vs processed) as well as the choice of features on the performance of various classifiers. The empirical results showed that the machine learning algorithms with strong regularization capabilities always outperformed others if provided with the most complete features set extracted from the raw accelerometer signals.

Based on the hypothesis that for accelerometer based human activity data, the most discriminative information could be found at multiple scales, a Multi-Scale Stacking Model (MSSM), which classifies the time series based on the features extracted at multiple scales, was proposed. The MSSM was designed for the time series with repetitive patterns, such as the accelerometer based human activity data. Along with two baseline models: the Multi-Scale Union Feature Model (MSUM) and the nearest neighbour model, it was evaluated on six time series datasets. Three of them were accelerometer based human activity data, and the rest were different types of time series data chosen from other domains. The empirical results indicated a strong advantage of the MSSM over other models on accelerometer based human activity data. Further analysis on the results confirmed the hypothesis that features from some scales were more discriminative than others.

Major Advisor: Weng-Keen Wong
Committee: Stewart Trost
Committee: Alan Fern
GCR: Ken Funk 

Kelley Engineering Center (campus map)
Shannon Thompson
1 541 737 7234
shannon.thompson at oregonstate.edu
Sch Elect Engr/Comp Sci
This event appears on the following calendars: