第3期 吕卓纹，等：加权CCA多信息融合的步态表征方法 ·453· 表4生成各步态特征时间对比 machine intelligence,2006,28(2):316-322 Table 4 Comparisons of generation time of gait character- [8]ZHANG Erhu,ZHAO Yongwei,XIONG Wei.Active en- istics ergy image plus 2DLPP for gait recognition[J].Signal pro- 算法 GEI GFI CFR PFR cessing,.2010,90(7):2295-2302. 时间s 5.94 31.87 47.6 47.5 [9]BASHIR K,XIANG Tao,GONG Shaogang.Gait recogni- 速率/fs 760 142 95 95 tion without subject cooperation[J].Pattern recognition let- ters,2010,31(13):2052-2060 [10]LAMT H W.CHEUNG K H.LIU JNK.Gait flow im- 4结束语 age:a silhouette-based gait representation for human 本文提出了一种新的步态表征方法。该方法 identification[J].Pattern recognition,2011,44(4): 增强了GI时序信息的表达，采用CCA去除特征 973-987. 间的冗余信息的同时将多通道信息融合成单通 [11]LEE C P,TAN A WC,TAN S C.Gait probability image: 道，使得新特征具有更丰富有益识别的信息，在 an information-theoretic model of gait representation[J] Journal of visual communication and image representa 协变量变化较大的USF数据集进行实验，由实验 tion,2014,25(6):1489-1492. 数据可知，本文提出的方法使得识别率得到了显 [12]DENG Muqing,WANG Cong,ZHENG Tongjia.Indi- 著的提升。 vidual identification using a gait dynamics graph[J].Pat- 下一步打算对本文算法直接选取已有的类能 tern recognition,2018,83:287-298. 量图分配到RGB通道；为了降低计算复杂度，可 [13]陈实，马天骏，黄万红，等.用于步态识别的多层窗口图 将融合算法CCA方法扩展到非线性耦合度量学 像矩U.电子与信息学报，2009,31(1)：116-119. 习空间。 CHEN Shi,MA Tianjun,HUANG Wanhong,et al.A multi-layer windows method of moments for gait recogni- 参考文献： tion[J].Journal of electronics and information technology, [1]CHEN Xin,WENG Jian,LU Wei,et al.Multi-gait recog- 2009.31(1)116-119. nition based on attribute discovery[J].IEEE transactions on [14]HOFMANN M.GEIGER J.BACHMANN S.et al.The pattern analysis and machine intelligence,2018,40(7): TUM gait from audio,image and depth (GAID)database: 1697-1710 Multimodal recognition of subjects and traits[J].Journal [2]GADALETA M.ROSSI M.IDNet:smartphone-based gait of visual communication and image representation,2014, 25(1):195-206 recognition with convolutional neural networks[J].Pattern [15]XING Xianglei,WANG Kejun,YAN Tao,et al.Com- recognition,2018,74:25-37. [3]BEN Xianye,ZHANG Peng,LAI Zhihui,et al.A general plete canonical correlation analysis with application to multi-view gait recognition[J].Pattern recognition,2016, tensor representation framework for cross-view gait recog- 50:107-117 nition[J].Pattern recognition,2019,90:87-98. [16]GAO Lei,QI Lin,CHEN Enqing,et al.Discriminative [4]EL-ALFY H,MITSUGAMI I,YAGI Y.Gait recognition multiple canonical correlation analysis for information fu- based on normal distance maps[J].IEEE transactions on sion[J].IEEE transactions on image processing,2018, cybernetics,.2018,48(5):1526-1539. 27(4):1951-1965. [5]AGGARWAL H.VISHWAKARMA D K.Covariate con- [17]石强，张斌，陈喆，等.异质影像融合研究现状及趋势 scious approach for Gait recognition based upon Zernike 自动化学报，2014.40(3)：385-396. moment invariants[J].IEEE transactions on cognitive and SHI Qiang,ZHANG Bin,CHEN Zhe,et al.Fusion tech- developmental systems,2018,10(2):397-407 niques for heterogeneous images:a survey[J].Acta auto- [6]何逸炜，张军平.步态识别的深度学习：综述.模式识 matica sinica2014.40(3:385-396. 别与人工智能，2018,31(5)：442-452 [18]WANG Chen,ZHANG Junping,WANG Liang,et al.Hu- HE Yiwei,ZHANG Junping.Deep learning for gait recog- man identification using temporal information preserving nition:a survey[J].Pattern recognition and artificial intelli- gait template[J].IEEE transactions on pattern analysis and gence,2018.31(5):442-452 machine intelligence,2012,34(11):2164-2176. [7]HAN Ju,BHANU B.Individual recognition using gait en- [19]SUN Shiliang.A survey of multi-view machine ergy image[J].IEEE transactions on pattern analysis and learning[J].Neural computing and applications,2013,表 4 生成各步态特征时间对比 Table 4 Comparisons of generation time of gait character￾istics 算法 GEI GFI CFR PFR 时间/s 5.94 31.87 47.6 47.5 速率/(f·s−1) 760 142 95 95 4 结束语 本文提出了一种新的步态表征方法。该方法 增强了 GFI 时序信息的表达，采用 CCA 去除特征 间的冗余信息的同时将多通道信息融合成单通 道，使得新特征具有更丰富有益识别的信息，在 协变量变化较大的 USF 数据集进行实验，由实验 数据可知，本文提出的方法使得识别率得到了显 著的提升。 下一步打算对本文算法直接选取已有的类能 量图分配到 RGB 通道；为了降低计算复杂度，可 将融合算法 CCA 方法扩展到非线性耦合度量学 习空间。 参考文献： CHEN Xin, WENG Jian, LU Wei, et al. Multi-gait recog￾nition based on attribute discovery[J]. IEEE transactions on pattern analysis and machine intelligence, 2018, 40(7): 1697–1710. [1] GADALETA M, ROSSI M. IDNet: smartphone-based gait recognition with convolutional neural networks[J]. Pattern recognition, 2018, 74: 25–37. [2] BEN Xianye, ZHANG Peng, LAI Zhihui, et al. A general tensor representation framework for cross-view gait recog￾nition[J]. Pattern recognition, 2019, 90: 87–98. [3] EL-ALFY H, MITSUGAMI I, YAGI Y. Gait recognition based on normal distance maps[J]. IEEE transactions on cybernetics, 2018, 48(5): 1526–1539. [4] AGGARWAL H, VISHWAKARMA D K. Covariate con￾scious approach for Gait recognition based upon Zernike moment invariants[J]. IEEE transactions on cognitive and developmental systems, 2018, 10(2): 397–407. [5] 何逸炜, 张军平. 步态识别的深度学习: 综述[J]. 模式识 别与人工智能, 2018, 31(5): 442–452. HE Yiwei, ZHANG Junping. Deep learning for gait recog￾nition: a survey[J]. Pattern recognition and artificial intelli￾gence, 2018, 31(5): 442–452. [6] HAN Ju, BHANU B. Individual recognition using gait en￾ergy image[J]. IEEE transactions on pattern analysis and [7] machine intelligence, 2006, 28(2): 316–322. ZHANG Erhu, ZHAO Yongwei, XIONG Wei. Active en￾ergy image plus 2DLPP for gait recognition[J]. Signal pro￾cessing, 2010, 90(7): 2295–2302. [8] BASHIR K, XIANG Tao, GONG Shaogang. Gait recogni￾tion without subject cooperation[J]. Pattern recognition let￾ters, 2010, 31(13): 2052–2060. [9] LAM T H W, CHEUNG K H, LIU J N K. Gait flow im￾age: a silhouette-based gait representation for human identification[J]. Pattern recognition, 2011, 44(4): 973–987. [10] LEE C P, TAN A W C, TAN S C. Gait probability image: an information-theoretic model of gait representation[J]. Journal of visual communication and image representa￾tion, 2014, 25(6): 1489–1492. [11] DENG Muqing, WANG Cong, ZHENG Tongjia. Indi￾vidual identification using a gait dynamics graph[J]. Pat￾tern recognition, 2018, 83: 287–298. [12] 陈实, 马天骏, 黄万红, 等. 用于步态识别的多层窗口图 像矩[J]. 电子与信息学报, 2009, 31(1): 116–119. CHEN Shi, MA Tianjun, HUANG Wanhong, et al. A multi-layer windows method of moments for gait recogni￾tion[J]. Journal of electronics and information technology, 2009, 31(1): 116–119. [13] HOFMANN M, GEIGER J, BACHMANN S, et al. The TUM gait from audio, image and depth (GAID) database: Multimodal recognition of subjects and traits[J]. Journal of visual communication and image representation, 2014, 25(1): 195–206. [14] XING Xianglei, WANG Kejun, YAN Tao, et al. Com￾plete canonical correlation analysis with application to multi-view gait recognition[J]. Pattern recognition, 2016, 50: 107–117. [15] GAO Lei, QI Lin, CHEN Enqing, et al. Discriminative multiple canonical correlation analysis for information fu￾sion[J]. IEEE transactions on image processing, 2018, 27(4): 1951–1965. [16] 石强, 张斌, 陈喆, 等. 异质影像融合研究现状及趋势[J]. 自动化学报, 2014, 40(3): 385–396. SHI Qiang, ZHANG Bin, CHEN Zhe, et al. Fusion tech￾niques for heterogeneous images: a survey[J]. Acta auto￾matica sinica, 2014, 40(3): 385–396. [17] WANG Chen, ZHANG Junping, WANG Liang, et al. Hu￾man identification using temporal information preserving gait template[J]. IEEE transactions on pattern analysis and machine intelligence, 2012, 34(11): 2164–2176. [18] SUN Shiliang. A survey of multi-view machine learning[J]. Neural computing and applications, 2013, [19] 第 3 期 吕卓纹，等：加权 CCA 多信息融合的步态表征方法 ·453·