簡易檢索 / 詳目顯示

回結果列表
研究生： 魏楚蒨
Wei, Chu-Chien
論文名稱： 使用骨架序列的走路姿態之身份辨識
Person Identification by Walking Gesture Using Skeleton Sequences
指導教授： 張世杰
Chang, Shih-Chieh
口試委員: 陳永昇
Chen, Yong-Sheng
洪樂文
Hong, Yao-Win
學位類別： 碩士
Master
系所名稱： 電機資訊學院 - 資訊工程學系
Computer Science
論文出版年： 2019
畢業學年度： 107
語文別： 英文
論文頁數： 25
中文關鍵詞： 身份辨識 、走路姿態 、骨架序列 、長短期記憶 、注意力機制
外文關鍵詞： person identification, walking gesture, skeleton sequence, long short-term memory, attention mechanism
相關次數： 點閱：2  下載：0 
分享至:
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報

    • 在身份辨識的問題中，大多數方法直接使用RGB影像來做辨識，但是這些方法忽略了照明和服裝的變化，因此這些方法只能在短時間內使用，在現實世界中，大多數時候我們希望系統能夠長時間使用，有研究顯示，人類的步態是獨一無二的，因此我們提出了一種根據行走姿勢來辨識人類身份的方法。我們想要利用骨架序列的資訊來找出人類行走姿勢的特徵，為了分析關節在空間上的關係，我們將三維座標中的骨架轉換為關節之間的距離和角度，然後我們使用雙向長短期記憶來探索骨架序列在時間上的關係，實驗結果證明我們的方法在用於長期觀察的數據集上比其他方法具有更好的表現。

    Among person identification problem, most methods use RGB images as input data.However, these methods ignore the changes of illumination and the different clothing, sothey can only be used in a short period. In the real world, most of the time we want the system to be used for a long time. Research shows that gait of a human is unique, so we propose a method to identify humans according to their walking gesture.In order to find out the characteristic of the individual walking gesture, we are interested in using the skeletal information. To analyze the spatial relationship of joints, we trans-form the skeleton in 3D coordinate into the distances and angles between joints. Then we use bidirectional long short-term memory to explore the temporal information of the skeleton sequence. On the datasets used for long term observation, experimental results show our method has better performance than other models.

    1 Introduction................................. 1 2 Preliminary 2.1 Related Works............................. 4 2.2 Motivation................................ 6 2.3 Datasets: BIWI & IAS-Lab.................. 7 3 Methodoloogy 3.1 Feature Extraction from Skeleton.......... 11 3.2 Long Short-Term Memory.................... 14 3.3 Attention Mechanism....................... 17 4 Experimental Results 4.1 Experimental Setup........................ 19 4.2 Compared Model............................ 20 4.3 Result Summary............................ 21 5 Conclusions.................................. 23 References .................................. 24

    [1] A. Elaoud, W. Barhoumi, H. Drira, and E. Zagrouba. Analysis of Skeletal ShapeTrajectories for Person Re-Identification. In J. Blanc-Talon, R. Penne, W. Philips,D. Popescu, and P. Scheunders, editors,ACIVS 2017 - 18th International Conferenceon Advanced Concepts for Intelligent Vision Systems, volume 10617 ofAdvancedConcepts for Intelligent Vision Systems 18th International Conference, ACIVS 2017,pages 138–149, Anvers, Belgium, Sept. 2017. Springer.
    [2] A. Haque, A. Alahi, and F. F. Li. Recurrent attention models for depth-based personidentification. 11 2016.
    [3] Liang Wang, Tieniu Tan, Huazhong Ning, and Weiming Hu. Silhouette analysis-based gait recognition for human identification.IEEE Transactions on Pattern Anal-ysis and Machine Intelligence, 25(12):1505–1518, Dec 2003.
    [4] M. Munaro, A. Basso, A. Fossati, L. Van Gool, and E. Menegatti. 3d reconstructionof freely moving persons for re-identification with a depth sensor. In2014 IEEE In-24
    ternational Conference on Robotics and Automation (ICRA), pages 4512–4519, May2014.
    [5] C. Prakash, R. Kumar, and N. Mittal. Recent developments in human gait research:parameters, approaches, applications, machine learning techniques, datasets and chal-lenges.Artificial Intelligence Review, 49(1):1–40, Jan 2018.
    [6] M. Schuster and K. Paliwal. Bidirectional recurrent neural networks.Trans. Sig.Proc., 45(11):2673–2681, Nov. 1997.

    QR CODE