针对步态识别问题,研究了受限玻尔兹曼机的步态特征提取及其识别.主要基于行人图像序列,通过背景分割、归一化以及步态周期的计算,采用叠加方法生成步态能量图,以此作为步态的特征图像,利用受限玻尔兹曼机自动获取步态特征.选取中科院CASIA步态数据库以及支持向量机、孪生支持向量机、神经网络与K-近邻方法对使用受限玻尔兹曼机方法的特征提取进行了研究,同时与主成分分析PCA、线性判别分析LDA、卷积神经网络CNN特征提取与识别方法进行了比较.
Aiming at gait recognition, the gait feature extraction and recognition of restricted Boltzmann machine(RBM)are studied. Based on the human pedestrian image sequence, the gait energy map is generated by background segmentation, normalization and gait cycle calculation. The gait energy map is used as the feature image of gait, and the gait feature is automatically acquired by RBM. The gait database of Chinese Academy of Sciences CASIA is chosen to study the gait recognition using RBM method for different classification methods, including support vector machine, twin support vector machine, neural network and K-nearest neighbor method. At the same time, this approach is compared with PCA, LDA and CNN feature extraction and recognition.
[1] NIYOGI S A, ADELSON E H. Analyzing gait with spatiotemporal surfaces[Z].IEEE Workshop on Motion of Non-rigid and Articulated Objects, Austin, TX, 1994. DOI:10.1109/mnrao.1994.346253.
[2] HUANG X X, BOULGOURIS N V. Gait recognition with shifted energy image and structural feature extraction[J]. IEEE Transactions on Image Processing, 2012, 21(4): 2256-2268. DOI:10.1109/tip.2011.2180914.
[3] CHOUDHURY S D, TJAHJADI T. Silhouette-based gait recognition using Procrustes shape analysis and elliptic Fourier descriptors[J]. Pattern Recognition, 2012, 45(9): 3414-3426. DOI: 10.1016/j.patcog.2012.02.032.
[4] BODER R, DRENNER A, FEHR D, et al. View-independent human motion classification using image-based reconstruction[J]. Image and Vision Computing, 2009, 27(8): 1194-1206. DOI: 10.1016/j.imavis.2008.11.008.
[5] CASTRO F M, MARÌN-JIMÈNEZ M J, MATA N G, et al. Fisher motion descriptor for multiview gait recognition[J]. International Journal of Pattern Recognition and Artificial Intelligence, 2017, 31(1): 1-40.DOI:10.1142/s021800141756002x.
[6] ZHENG S, ZHANG J G, HUANG K Q, et al. Robust view transformation model for gait recognition[Z]. 18th IEEE International Conference on Image Processing, Brussels, Belgium, 2011. DOI:10.1109/icip.2011.6115889.
[7] YU S Q, CHEN H F, WANG Q, et al. Invariant feature extraction for gait recognition using only one uniform model[J]. Neurocomputing, 2017, 239: 81-93. DOI: 10.1016/j.neucom.2017.02.006.
[8] ZHANG L F, ZHANG L P, TAO D C, et al. A sparse and discriminative tensor to vector projection for human gait feature representation[J]. Signal Processing, 2015, 106:245-252. DOI: 10.1016/j.sigpro.2014.08.005.
[9] ROY A, SURAL S, MUKHERJEE J. Gait recognition using pose kinematics and pose energy image[J]. Signal Processing, 2012, 92(3):780-792. DOI: 10.1016/j.sigpro.2011.09.022.
[10] GUAN Y, LI C T, HU Y. Random subspace method for gait recognition[Z]. IEEE International Conference on Multimedia & Expo Workshops., Melbourne, Australia, 2012. DOI:10.1109/icmew.2012.55 2012:284-289.
[11] XING X L, WANG K J, YAN T, et al. Complete canonical correlation analysis with application to multi-view gait recognition[J]. Pattern Recognition, 2016, 50: 107-117. DOI: 10.1016/j.patcog.2015.08.011.
[12] WEI L, KUO J, PENG J L. Gait recognition via GEI subspace projections and collaborative representation classification[J]. Neurocomputing, 2018, 275:1932-1945. DOI: 10.1016/j.neucom.2017.10.049.
[13] 郭欣,王蕾,宣伯凯,等.基于有监督Kohonen神经网络的步态识别[J].自动化学报,2017,43(3):430-438. DOI: 10.16383/j.aas.2017.c160114.
[14] 王科俊,丁欣楠,邢向磊,等.多视角步态识别综述[J].自动化学报,2019,45(5):841-852. DOI: 10.16383/j.aas.2018.c170599.
[15] MA G, K WANG Y, WU L G. Subspace ensemble learning via totally-corrective boosting for gait recognition[J]. Neurocomputing, 2017, 224:119-127. DOI: 10.1016/j.neucom.2016.10.047.
[16] DENG M Q, WANG C, CHENG F J, et al. Fusion of spatial-temporal and kinematic features for gait recognition with deterministic learning[J]. Pattern Recognition, 2017, 67:186-200. DOI: 10.1016/j.patcog.2017.02.014.
[17] DENG M Q, WANG C, CHEN Q F. Human gait recognition based on deterministic learning through multiple views fusion.[J]. Pattern Recognition Letters, 2016, 78(C):56-63. DOI: 10.1016/j.patrec.2016.04.004.
[18] HINTON G E, OSINDERO S, TEH Y W. A fast learning algorithm for deep belief nets[J]. Neural Computation, 2006, 18(7):1527-1554. DOI: 10.1162/neco.2006.18.7.1527.
[19] ROUX N, BENGIO Y. Representational power of restricted boltzmann machines and deep belief networks[J]. Neural Computation, 2008, 20(6):1631-1649. DOI: 10.1162/neco.2008.04-07-510.
[20] HINTON G E. Training products of experts by minimizing contrastive divergence[J]. Neural Computation, 2002, 14(8): 1771-1800. DOI:10.1162/089976602760128018.
[21] 刘丽丽.基于最外轮廓的步态识别研究[D].济南:山东大学,2012.
[22] WANG L, TAN T, NING H, et al. Silhouette analysis-based gait recognition for human identification[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2003, 25(12):0-1518. DOI:10, 1109/tpami. 2003. 1251144.
