基于Laguerre正交基神经网络的动态手势识别*

南京大学学报(自然科学版) ›› 2011, Vol. 47 ›› Issue (5): 515–523.

基于Laguerre正交基神经网络的动态手势识别*

李文生^1**解梅^1,2，姚琼¹

出版日期:2015-04-24 发布日期:2015-04-24
作者简介: 1.电子科技大学中山学院计算机学院，中山，528402; 2.电子科技大学电子工程学院，成都,610054
基金资助:
广东省自然科学基金(8152840301000009)，广东省科技计划(20098030803031)

Dynamic gesture recognition based on Laguerre orthogonal basis neural network

Li Wen-Shenh ¹，Xie Mei ^1,2，Yao Qiong¹

Online:2015-04-24 Published:2015-04-24
About author: (1 .School of Computer, University of Electronic Science and Technology of China, Ghongshan Institute,
Zhongshan, 528402，China;2. School of Electronic Engineering, University of Electronic Science
and Technologv of China, Chengdu，61005，China)

摘要/Abstract

摘要： 提出一种基于Lagucrre正交基前向神经网络的动态手势识别方法.首先根据多项式逼近和矩阵理论，构造了一种以Lagucrre正交多项式作为隐含层神经元激励函数的多输入、多输出三层前向神
经网络模型，在网络权值迭代计算公式基础上推出一种基于伪逆的直接计算网络权值方法，避免求取权值的反复迭代过程;提出一种快速的基于颜色的指尖检测跟踪算法以便实时获取指尖运动轨迹，并提取
指尖运动轨迹的特征向量作为Lagucrre神经网络的输入向量;通过预先获取的动态手势样木(包括手势输入向量和预期结果)训练Lagucrre神经网络，利用经过训练的Laguerre神经网络来识别通过摄像头获
取的动态手势.测试结果表明:Lagucrre正交基前向神经网络能够提高学习训练速度和精度，而且在动态手势识别方面具有较好的鲁棒性和泛化能力，具有较高的识别准确率.

Abstract: As a powerful tool for learning highly nonlinear system, BP neural networks are widely used in the field of nonlinear system identification such as pattern recognition. However, due to its shortcomings such as slow
convergence rate, the danger of over fitting and low recognition accuracy, the traditional BP neural networks perform poorly in dynamic gesture recognition, especially for online dynamic gesture training, In this paper, a novel
method of dynamic gesture learning and recognition based on Lagucrre orthogonal basis feed-forward neural network was put forward. First，based on polynomial interpolation and matrix theory, a three-laycr MlMO feed-forward
neural network which hidden layer neurons arc activated by a group of Lagucrre orthogonal polynomial functions was constructed. In our research, the weights-updating formula for the neural networks was derived by adopting the
standard BP training method，and then a pseudo-inverse based method which could determine the network weights directly without lengthy iterative training was proposed. Second，a rapid algorithm to detect and track the fingertips
in real time based on machine vision was proposed. At first，a 2D color histogram on HSV color space was computed as the result of online training of fingertip colors，and it was used as the basis of fingertip detection and tracking;
丁hen color probability distribution I(x，y) of the target was calculated through back-projection on the 2D color histogram and transformed into a binary image PCx，y) according to a given threshold;Then the target region of
fingertips was determinated through edge detection and the centroid positions of targets could be calculated;At last，the trajectories of fingertips could be acquired by tracking the centroids of the fingertips through MDF(Minimum
Distance First) and characteristic vectors of the dynamic gestures could be obtained. Third，some pre-obtained dynamic gestures that include characteristic vectors of gestrucs and expected outputs were used as sample gestures to
train the Lagucrre neural networks, and new dynamic gestures acquired by a camera can be recognized through the trained Lagucrre neural networks. Experiment results show that Lagucrre orthogonal basis feed-forward neural
network can enhance the speed and precision of network training, and improve the robust and accuracy of dynamic gesture recognition.

李文生^1**解梅^1,2，姚琼¹. 基于Laguerre正交基神经网络的动态手势识别*[J]. 南京大学学报(自然科学版), 2011, 47(5): 515–523.

Li Wen-Shenh ¹，Xie Mei ^1,2，Yao Qiong¹
. Dynamic gesture recognition based on Laguerre orthogonal basis neural network
[J]. Journal of Nanjing University(Natural Sciences), 2011, 47(5): 515–523.

参考文献

[1]Alexander T C, Ahmcd H S, Anagnostopoulos G C. An open source framework for real-time, incremental，static and dynamic hand gesture learning and recognition. Jacko J A. Human}omputer interaction, Part II, HCII 2009. Berlin;Springer-Verlag, 2009，X611:123一130.
[2]Jose M M, Moya, Alvaro A, et al. Low-cost gesture-based interaction for intelligent environments. Omatu S.The 10th international Work Conference on Artificial Neural Networks, Part II. Bcrlin;Springer-Verlag, 2009，5518:752一755.
[3]Hardy F,Javier R-de-S,Rodrigo V.Real-timehand gesture detection and recognition using boosted classifiers and active learning. Pirhonen A, Brcwstcr S. 2007 IEEE PacifirRim Symposium on Image and Video Technology. Berlin: Springer-Verlag,2007,4872:533~547
[4]Kaustubh S. Hand gesture modelling and recognition involving changing shapes and trajectories, using a predictive eigen tracker. Pattern Recognition Letters, 2007，28; 329一334.
[5]Jiyoung P, Juneho Y. Efficient fingertip tracking and mousepointer control for a human mouse.Crowley J L.The 3rd International Conference on Computer Vision Systems. Berlin;Springer-Verlag,2003,2323:88~97
[6]He Z J，Yu W D, Chen Z Q, et al. A BP neural network algorithm based on genetic algorithms and its application. Journal of Nanjing University(Natural Sciences)，2003,39(5): 459-465.
(土崇骏，于汉粉滌 ,陈兆乾等.一种基于遗传算法的神经网络算法及其应用.南京大学学报(自然科学),2003,39C5):459一465)
[7]Martin T H, Howard B D, Mark H B. Neural network design. Beijing; China Machine Press, 2002, 197一257.
[8]Yang D H, Nian X H, Yang S Y.Two kinds of improved BP learning algorithms for neural network. Journal of Changsha University, 2004，18(4): 54一57.(杨东侯，年晓红，杨胜跃.两种改进的YP神经网络学习算法.长沙大学学报，2001，18(4):54一57).
[9]Zhou Y, Xu B L. Orthogonal method for training neural networks. Journal of Nanjing Univcrsity(Natural Scicnccs),2001,37(1);72一78 (周毅，徐柏龄.神经网络中的正交设计法研究. 南京大学学报(自然科学)，2001,37(1) : 72一78.
[10]Wu X J，Wang S T，Yang J Y.The study on the orthogonal polynomials-based neural networks and its properties. Computer Engineering and Application, 2002 , 38 (9 ) ; 25一26.(吴小俊，王士
同，杨静宇.基于正交多项式函数的神经网络及其性质研究.计算机工程与应用，2002,38(9);25 ~26).
[11]Jagdish C P, Alex C K. Nonlinear dynamic system identification using Chebyshev functional link artificial neural networks, IEEE Transactions on Systems, N1an, and Cybernetics, Part B; Cybernetics, 2002，32(4));505一511.
[12Li M, He Y G Nonlinear system identification using adaptive Chebyshev neural networks, intelligent Computing and intelligent Systems, 2010; 243 ~247.
[13]Zhang Y N, Chen Y L, Jiang X H et al aWeights-directly-detemined and Structnre adaptively tuned neural network based on Chebyshev basis functions. Computer Science 2009, 36(6): 210一213.(张雨浓，陈裕降，姜孝华
等.一种权值直接确定及结构自适应的Chcbyshcv基函数神经网络.计算机科学，2009, 36 (6):210一213).
[14]Zhang Y N, Xiao X C，Chcn Y W，et al. Number determination of hidderrlaycr nodes for Hermite feed-forward neural network. Journal of Zhcjiang Univcrsity(Enginccring Science). 2010, 44(2); 271一275.(张雨浓，
肖秀春，陈扬文等.H ermite前向神经网络隐节点数目自动确定.浙江大学学报(工学版)，2010，44(2):271一275).
[15]Zhang Y N, Zhong T K，Li W, et al. Laguerre orthogonal basis feed-forward neural network with its weights determined directly. Journal of Jinan University( Natural Science and Medicine Edition)，2008，29(3) ;219一263.
(张雨浓，钟童科，李巍等.Lagucrrc正交基前向神经网络及其权值直接确定法.暨南大学学报(自然科学),2008,29(3);249一253).

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed