基於深度神經網路的手勢辨識研究

以作者查詢圖書館館藏

、以作者查詢臺灣博碩士

、以作者查詢全國書目

、勘誤回報

、線上人數：17

、訪客IP：3.138.113.188

姓名

蔡緯豐(Wei-Feng Tsai) 查詢紙本館藏

畢業系所

電機工程學系

論文名稱

基於深度神經網路的手勢辨識研究
(Hand Gesture Recognition Based on Deep Neural Network)

相關論文

★ 人體姿勢矯正項鍊配載影像辨識自動校準及手機接收警告系統	★ 以遞迴式神經網路補償模型預測控制於永磁同步馬達定位
★ 基於卷積神經網路及色彩影像技術之火焰辨識

檔案

[Endnote RIS 格式]

[Bibtex 格式]

[相關文章]

[文章引用]

[完整記錄]

[館藏目錄]

至系統瀏覽論文 ( 永不開放)

摘要(中)

本文的目標是要實現使用網路攝影機即時追蹤影像範圍內的手部區域並且辨識手勢，應用於家電控制與人機互動等領域。我們首先利用膚色檢測和形態學處理分離影像，去除不必要的訊息，再利用背景相減法抓取手部的位置的區域ROI(Region Of Interest)。接著，為了避免雜訊影響到手部區塊，我們使用KCF(Kernelized Correlation Filters)演算法追蹤偵測到的手部區域ROI。最後將ROI的大小調整到100 * 120的大小，再將圖像輸入CNN (Convolutional Neural Networks)網路中進行多種手勢的辨識。接著重複上述追蹤和辨識的步驟達到即時的效果。本研究使用參考Alexnet和VGGnet網路的兩種架構進行訓練和比較，最後在訓練數據集中達到99.9%的辨識率，測試數據集有95.61%的辨識率。

摘要(英)

The purposes of this paper are to achieve hand gesture recognition and tracking hand position in real time via web camera. First, using skin-color detect and morphological operations to remove unnecessary noise. Then use the background subtraction method to determine the ROI(Region Of Intereest) region of hand. After obtaining the hand region, Kernel Correlation Filters (KCF) algorithm is used to track the hand. Finally, the hand area is scaled to the size of 100 * 120, then the fixed size of the image input to our CNN (Convolutional Neural Networks) network for identification, in order to achieve the effect of identifying a variety of gestures. And repeat tracking and identification to achieve the real time performance.This research used two frameworks which referenced Alexnet and VGGnet for training and comparison. Finlly, a 99.9% recognition rate is achieved in the training data. The test data set has a recognition rate of 95.61%.

關鍵字(中)

★ 手部偵測
★ KCF追蹤
★ CNN
★ 手勢辨識

關鍵字(英)

★ hand detection,
★ KCF tracking
★ CNN
★ gesture recognition

論文目次

中文摘要 iv
英文摘要 v
致謝 vi
目錄 vii
圖目錄 x
表目錄 xiii
第一章緒論 - 1 -
1-1 簡介 - 1 -
1-2 文獻回顧 - 2 -
1-3 研究動機與方法 - 3 -
1-4 主要貢獻 - 5 -
1-5 論文架構 - 5 -
第二章系統描述 - 6 -
2-1 硬體 - 6 -
2-2 使用軟體 - 7 -
2-3 系統架構 - 9 -
第三章手部偵測與追蹤 - 10 -
3-1 各種色彩空間介紹 - 11 -
3-1-1 RGB色彩空間 - 11 -
3-1-2 HSV色彩空間 - 11 -
3-1-3 YCbCr色彩空間 - 13 -
3-2 膚色切割法 - 13 -
3-3 形態學處理與平滑 - 16 -
3-3-1 侵蝕(Erosion) - 17 -
3-3-2 膨脹(Dilation) - 18 -
3-3-3 斷開(Opening) - 19 -
3-3-4 閉合(Closing) - 20 -
3-3-5 平滑處理 - 21 -
3-3-6 手部雜訊處理 - 22 -
3-3-7 背景相減法 - 24 -
3-4 手部追蹤 - 26 -
3-4-1 Linear regression - 26 -
3-4-2 Cyclic shift - 27 -
3-4-3 Kernal - 30 -
3-4-4 Fast detection - 31 -
3-4-5 Kernel Correlation - 32 -
3-4-6 追蹤流程 - 33 -
第四章手勢辨識 - 35 -
4-1 卷積神經網路 - 36 -
4-1-1 卷積層(Convolutional layer) - 37 -
4-1-2 線性整流層（Rectified Linear Units layer, ReLU layer） - 40 -
4-1-3 池化層( Pooling Layer ) - 41 -
4-1-4 全連接層( Full connected layer ) - 42 -
4-2辨識架構 - 43 -
4-3訓練方法 - 46 -
第五章實驗結果與討論 - 50 -
5-1實驗流程 - 51 -
5-2 訓練最加化 - 52 -
5-3 網路架構與結果 - 58 -
5-4 討論 - 64 -
第六章結論與建議 - 66 -
參考文獻 - 67 -

參考文獻

[1] S. Suzuki, “Topological structural analysis of digitized binary images
by border following. Computer vision, graphics ”, and image processing, 30(1): 32-46 , 1985.
[2] J. Sklansky, “Finding the convex hull of a simple polygon. Pattern
Recognition Letters ”, 1(2): 79-83 , 1982.
[3] Srinivas Ganapathyraju, “ Hand gesture recognition using convexity hull defects to control an industrial robot ”, Instrumentation Control and Automation (ICA), 28-30 Aug, 2013.
[4] Microsoft, “Kinect for Windows”, http://www.kinectforwindows.org/,
2012.
[5] A. Memo, L. Minto, P. Zanuttigh, “Exploiting Silhouette Descriptors and Synthetic Data for Hand Gesture Recognition ”, 2015.
[6] Cem Keskin, Furkan K?rac, Yunus Emre Kara, Lale Akarun, “Real time hand pose estimation using depth sensors ”, IEEE International Conference on Computer Vision Workshops, pp. 6-13 Nov, 2011.
[7] Leo Breiman , “Random Forests ”, Journal Machine Learning pp. 5-32 Oct, 2001.
[8] Trong-Nguyen Nguyen, Duc-Hoang Vo, Huu-Hung Huynh, Jean Meunier, “Geometry-based static hand gesture recognition using support vector machine”, Control Automation Robotics & Vision (ICARCV), 10-12 Dec, 2014.
[9] T. Starner, A. Pentland, “ Real-time american sign language recognition from video using hidden markov models ”, Motion-Based
Recognition. Springer Netherlands, 227-243, 1997.
[10] A. Krizhevsky, I. Sutskever , G. E. Hinton, “ Imagenet classification
with deep convolutional neural networks ”, Advances in neural information processing systems, 1097-1105 , 2012.
[11] ImageNet Large Scale Visual Recognition Challenge (ILSVRC) , “ http://www.image-net.org/challenges/LSVRC/ ”, 2010 – 2017.
[12] Mengmeng Han, Jiajun Chen, Ling Li, Yuchun Chang,“Visual Hand Gesture Recognition with Convolution Neural Network”, 21 July , 2016.
[13] P. KadewTraKuPong, R. Bowden,“An improved adaptive background mixture model for real-time tracking with shadow detection”, 2001.
[14] Z. Zivkovic, F. van der Heijden ,“Efficient adaptive density estimation per image pixel for the task of background subtraction”, 2006.
[15] Joao F. Henriques, Rui Caseiro, Pedro Martins, and Jorge Batista, “High-speed tracking with kernelized correlation filters”, Pattern Analysis and Machine Intelligence, IEEE Transactions on, vol. 37, no. 3, pp. 583–596, 2015.
[16] K. Simonyan and A. Zisserman,“Very deep convolutional networks for large-scale image recognition ” , arXiv preprint arXiv:1409.1556, 2014.
[17] OpenCV,“https://opencv.org/ ”, 2000
[18] Tensorflow,“https://www.tensorflow.org/ ”, 2015
[19] Youlian Zhu, Cheng Huang and Jiajun Chen, “Face Detection Method Based on Multi-feature Fusion in YCbCr Color Space ”, International Congress on Image and Signal Processing, pp. 1249-1252, Oct. 2012.
[20] C. Tomasi ; R. Manduchi,“Bilateral filtering for gray and color images”, Computer Vision, Sixth International Conference on , pp. 7-7 Jan , 1998.
[21] Maycas Nadal, Catlos,“Input-Output Kernel Regression applied to proteinprotein interaction network inferenc”, PhD Thesis , 2015.
[22] Yan LeCun“LeNet-5, convolutional neural networks ”, http: //yann.lecun.com/exdb/lenet/, 1998.
[23] Sebastian Ruder, “ An overview of gradient descent optimization
Algorithms ” , 2016.
[24] Ning Qian, “On the momentum term in gradient descent learning algorithms. ”, Neural networks :the official journal of the International Neural Network Society, 12(1):145–151 , 1999.
[25] Yurii Nesterov, “A method for unconstrained convex minimization problem with the rate of convergence o(1/k2) ”, Doklady ANSSSR (translated as Soviet.Math.Docl.), 269:543–547 , 1993.
[26] John Duchi, Elad Hazan, Yoram Singer ,“ Adaptive Subgradient Methods for Online Learning and Stochastic Optimization ” , Journal of Machine Learning Research, 2121–2159, 2011.
[27] Matthew D. Zeiler ,“ ADADELTA: An Adaptive Learning Rate Method ” , 2012.
[28] Diederik P. Kingma , Jimmy Lei Ba , “ Adam: a Method for Stochastic Optimization ”, InternationalConference on Learning Representations, pp. 1–13 , 2015.
[29] Srivastava, N., Hinton, G., Krizhevsky, A., Sutskever, I., & Salakhutdinov, R , “ Dropout: A simple way to prevent neural networks from overfitting. ”, The Journal of Machine Learning Research, 15(1), 1929-1958 , 2014
[30] Geoffrey E. Hinton, Nitish Srivastava, Alex Krizhevsky, Ilya Sutskever, Ruslan R. Salakhutdinov , “ Improving neural networks by preventing co-adaptation of feature detectors ” , 2012.

指導教授

鍾鴻源莊堯棠(Hung-Yuan Chung Yau-Tarng Juang)

審核日期

2018-7-24

推文