| Summary: | 碩士 === 國立成功大學 === 電腦與通信工程研究所 === 93 === Gait, or the style of walking, has recently been a popular topic in vision-based analysis. It is believed that gait is unique to every individual and reflects body conditions. However, current vision-based works about gait are mostly devoted to the application of human recognition, and abnormal walking styles are not included for discussion. On the other hand, behavior analysis in home care system or other surveillance application is needed for detecting unusual events so as to making proper response.
Considering the above points, in this thesis, a vision-based method is proposed to analyze abnormal types of walking. With this analysis, a few abnormal types of walking can be distinguished. In the proposed method, a background subtraction algorithm is first applied to segment out the silhouette of the walker at each frame in a sequence. For each frame, we define a feature based on the length between two legs, called aspect ratio, and by observing this feature value across time (or frame), a periodic wave is obtained. Then, based on the quantities measured from this wave, several attributes about gait are extracted. Finally, a preliminary classification is used to determine if the style of walking in a sequence is among the abnormal types we can find.
|
|---|
