運用Kinect利用步態做人的辨識

Gait pattern for person recognition, an Application of Kinect

黃柏豪(Bo-Hao Huang)

指導教授 ： 荊宇泰

國立交通大學/資訊學院/生醫工程研究所/碩士(2015年)

https://doi.org/10.6842/NCTU.2015.00736

摘要

身份辨識在場景監控、安全管理以及門禁系統上扮演重要的角色。一般是以密碼或是識別證來確認身分。但是會有遺失或是被盜用的風險在。而生物辨識技術是以個人生物特徵來進行辨識。生物特徵具有其生物個體的獨特性，所以利用在身分辨識上更具安全性。本論文研究人的行走模式中，能夠當作辨識身分的特徵。
本研究的實驗方法是利用Kinect感測器捕捉受測者在跑步機上行走的骨架，並以運動學的角度分析，希望能夠找出個人特徵。從骨架資料中計算出關節角度，並分析頻譜資料，可以發現同一受測者在不同日期所蒐集的資料間，會有相同的頻率以及趨勢存在；而不同受測者之間，可以找出每位受測者都有與其他受測者資料不同的頻率。因此選用關節角度曲線的頻率，和其他體型特徵，當成特徵數列。
決策樹分析常用於分類或是預測模型的方法，透過樹狀結構的規則所構成的樹狀圖，讓資料能夠進行分類。決策樹結構簡單，分類規則容易理解，因此選用決策樹的模型來進行資料的分類。
實驗的結果說明本研究所選擇的特徵具有個人的獨特性以及可辨識性，而步態與個人體型以及走路習慣有關，不容易被模仿，因此步態辨識可說是相當可靠的生物辨識技術。

關鍵字

身分辨識 ； 生物特徵 ； 步態 ； Kinect ； 決策樹

並列摘要

Identity verification plays an important role in surveillance, security and access control system. Password and recognition ID card are available for use. But there are some risk like lost or stolen. Biometric techniques use human characteristics to identify. Biometric techniques have its individual, so it more safety. This research use human gait pattern to identify.
Experiment use Kinect sensor to capture participant’s walking skeleton on the treadmill. And analysis in kinematics way to find recognizable features. Calculate joint angles from skeleton data, and analysis in frequency domain. For each participant, angles has the same frequency and trend between different date data. Different participant has their individual frequency. Thus we choose joint angles’ frequency domain to be the identify features.
Decision tree is a decision support tool that uses a tree-like graph or model of decisions and their possible consequences. The structure is more simple and the classification rule is easy to understand.
Result shows our features have individuality and identifiability. And human gait has something to do with physique and walking habits. So gait pattern is difficult to be imitated. And we can say that gait recognition could be a reliable biometric technique.

並列關鍵字

Identity verification ； Biometric technique ； Gait ； Kinect ； Decision tree

參考文獻

1. L. Lee, W. E. L. Grimson, “Gait Analysis for Recognition and Classification”, Fifth IEEE International Conference, Automatic Face and Gesture Recognition, pp. 148-155, 2002.

4. Brook Galnaa, et al., “Accuracy of the Microsoft Kinect sensor for measuring movement in people with Parkinson's disease”, Gait & Posture 39, Volume 39, Issue 4, pp. 1062–1068, 2014.

5. N. A. Borghese ,L. Bianchi, F. Lacquaniti, “Kinematic determinants of human locomotion”, Journal of Physiology, pp.863-879, 1996.

6. Bouchrika, “Gait Analysis and Recognition for Automated Visual Surveillance”, University of Southampton, Department of Electronics and Computer Science, Doctoral Thesis, 2008.

7. Liu, Z., Sarkar, S., “Simplest representation yet for gait recognition: averaged silhouette”, 17th International Conference on (Volume:4), Pattern Recognition, pp.211-214, 2004.