Oregon State University

We’d like your feedback: Calendar User Survey – Event Creator Survey



Event Details

MS Final Examination – Michael Anderson

Thursday, June 13, 2013 1:00 PM - 3:00 PM

Physical Activity Recognition of Free-Living Accelerometer Data Using Change-Point Detection Algorithms and Hidden Markov Models
Physical activity recognition using accelerometer data is a rapidly emerging field with many real-world applications. Much of the previous work in this area has assumed that the accelerometer data has already been segmented into pure activities, and the activity recognition task has been to classify these segments. In reality, activity recognition would need to be applied to "free-living" data, which is collected over a long, continuous time period and would consist of a mixture of activities. In this thesis, we explore two approaches for segmenting realistic free-living time series data.

In the first approach, we apply a top-down strategy in which we segment free-living data using change-point detection algorithms and then classify the resulting segments using supervised learning techniques. In the second approach, we employ a bottom-up strategy in which we split the time series into small fixed-length windows, classify these windows, and then smooth the predictions using an HMM. Our results clearly show that the bottom-up approach is far superior to the top-down approach in both accuracy and timeliness of detection.

Major Advisor: Weng-Keen Wong
Committee: Prasad Tadepalli
Committee: Raviv Raich
GCR: Hector Vergara 

Kelley Engineering Center (campus map)
Nicole Thompson
1 541 737 3617
Nicole.Thompson at oregonstate.edu
Sch Elect Engr/Comp Sci
This event appears on the following calendars: