利用Capped核范数正则化的人体运动捕获数据恢复

胡文玉; 朱雪芳; 易云

doi:10.3724/SP.J.1089.2023.19486

利用Capped核范数正则化的人体运动捕获数据恢复

1 赣南师范大学数学与计算机科学学院赣州 341000;
2 江西环境工程职业学院通讯与信息学院赣州 341000

基金项目:

国家自然科学基金(62266002,61863001,61962003,61502107)

江西省自然科学基金(20224BAB202004,20202BAB202017).

详细信息

作者简介:
胡文玉(1982-),男,博士,教授,硕士生导师,CCF会员,主要研究方向为机器学习、计算机视觉,huwenyu@gnnu.edu.cn;朱雪芳(1997-),女,硕士,讲师,主要研究方向为机器学习;易云(1983-),男,博士,副教授,硕士生导师,主要研究方向为计算机视觉、视频内容理解.

中图分类号: TP391.41
计量
- 文章访问数: 90
- HTML全文浏览量: 2
- PDF下载量: 51
出版历程
- 收稿日期: 2021-11-07
- 修回日期: 2022-01-09
- 网络出版日期: 2023-11-06
- 刊出日期: 2023-08-19

Human Motion Capture Data Recovery Using Capped Nuclear Norm Regularization

1 School of Mathematics and Computer Science, Gannan Normal University, Ganzhou 341000;
2 School of Communications and Information, Jiangxi Environmental Engineering Vocational College, Ganzhou 341000

摘要

摘要: 结合人体运动数据的低秩性，将人体运动捕获数据恢复问题建模为低秩矩阵填充问题.不同于传统方法采用核范数作为矩阵秩函数的凸松弛，引入非凸的矩阵Capped核范数(CaNN).首先，建立基于CaNN正则化的人体运动捕获数据恢复模型;其次，利用交替方向乘子法，结合截断参数自适应学习与(逆)离散余弦傅里叶变换对模型进行快速求解;最后，在CMU数据集和HDM05数据集上，将CaNN模型与经典的TSMC，TrNN，IRNN-Lp和TSPN模型进行对比实验.恢复误差和视觉效果比较结果表明，CaNN能够有效地对失真数据进行恢复，且恢复后的运动序列与真实运动序列逼近度较高.
- 运动捕获 /
- 低秩结构 /
- 矩阵填充 /
- Capped核范数 /
- 交替方向乘子法
Abstract: Using the low-rank property of human motion data, the problem of recovering human motion capture data is modeled as a low-rank matrix completion problem. Different from the traditional methods which utilize the nuclear norm as the convex relaxation of rank function, a non-convex matrix Capped nuclear norm（CaNN）is introduced in this paper. Firstly, the recovery model of human motion capture data is established based on the CaNN regularization. Secondly, the model is efficiently solved by using the alternative direction method of multipliers, combined with adaptive learning for the truncated parameter and（inverse） discrete cosine Fourier transform. Finally, the proposed model CaNN is compared with four classical models, i.e., TSMC, TrNN,IRNN-Lp and TSPN, on CMU dataset and HDM05 dataset. By comparing the recovery error and visual effect,the experimental results show that CaNN has a good ability to recover the corrupted motion data, and the recovered motions can well approximate the true ones.
- motion capture /
- low-rank structure /
- matrix completion /
- Capped nuclear norm /
- alternating direction method of multipliers

HTML全文

参考文献(27)

[1]	Sun Huaijiang, Xia Guiyu, Zhang Guoqing, et al. Survey on reuse of human motion capture data[J]. Journal of Data Acquisition and Processing, 2017, 32(1): 1-16(in Chinese)
[2]	Wang Yaming, Lu Tao, Han Yonghua. Human motion capture data recovery based on skeleton constraint[J]. Computer Systems & Applications, 2018, 27(5): 17-25(in Chinese)
[3]	Malti A. On the exact recovery conditions of 3D human motion from 2D landmark motion with sparse articulated motion[J]. Computer Vision and Image Understanding, 2021, 202: 103072
[4]	Wangni J Q, Lin D H, Liu J, et al. Towards statistically provable geometric 3D human pose recovery[J]. SIAM Journal on Imaging Sciences, 2021, 14(1): 246-270
[5]	Yang J Y, Guo X, Li K, et al. Spatio-temporal reconstruction for 3D motion recovery[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2020, 30(6): 1583-1596
[6]	Rose C, Cohen M F, Bodenheimer B. Verbs and adverbs: multidimensional motion interpolation[J]. IEEE Computer Graphics and Applications, 1998, 18(5): 32-40
[7]	Aristidou A, Lasenby J. Real-time marker prediction and CoR estimation in optical motion capture[J]. The Visual Computer, 2013, 29(1): 7-26
[8]	Yin W J, Yin H, Kragic D, et al. Graph-based normalizing flow for human motion generation and reconstruction[OL]. [2021-11-08]. https://arxiv.org/abs/2104.03020
[9]	Mao C, Qin S F, Wright D. A sketch-based approach to human body modelling[J]. Computers & Graphics, 2009, 33(4): 521-541
[10]	Herda L, Fua P, Plankers R, et al. Skeleton-based motion capture for robust reconstruction of human motion[C] //Proceedings of Computer Animation. Los Alamitos: IEEE Computer Society Press, 2000: 77-83
[11]	Li L, McCann J, Pollard N S, et al. DynaMMo: mining and summarization of coevolving sequences with missing values[C] //Proceedings of the 15th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York: ACM Press, 2009: 507-516
[12]	Lai R Y Q, Yuen P C, Lee K K W. Motion capture data completion and denoising by singular value thresholding[C] //Proceedings of Eurographics Association. Aire-la-Ville: Eurographics Association Press, 2011: 45-48
[13]	Feng Y F, Xiao J, Zhuang Y T. Exploiting temporal stability and low-rank structure for motion capture data refinement[J]. Information Sciences, 2014, 277: 777-793
[14]	Hu Y, Zhang D B, Ye J P, et al. Fast and accurate matrix completion via truncated nuclear norm regularization[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013, 35(9): 2117-2130
[15]	Chen B J, Sun H J, Xia G Y, et al. Human motion recovery utilizing truncated schatten p-norm and kinematic constraints[J]. Information Sciences, 2018, 450: 89-108
[16]	Zhang X Q, Chen B J, Sun H J, et al. Robust low-rank kernel subspace clustering based on the schatten p-norm and correntropy[J]. IEEE Transactions on Knowledge and Data Engineering, 2020, 32(12): 2426-2437
[17]	Lu C Y, Tang J H, Yan S C, et al. Nonconvex nonsmooth low rank minimization via iteratively reweighted nuclear norm[J]. IEEE Transactions on Image Processing, 2016, 25(2): 829-839
[18]	Nie F P, Hu Z X, Li X L. Matrix completion based on non-convex low-rank approximation[J]. IEEE Transactions on Image Processing, 2019, 28(5): 2378-2388
[19]	Hu W Y, Wang Z, Liu S, et al. Motion capture data completion via truncated nuclear norm regularization[J]. IEEE Signal Processing Letters, 2018, 25(2): 258-262
[20]	Cui Q J, Chen B J, Sun H J. Nonlocal low-rank regularization for human motion recovery based on similarity analysis[J]. Information Sciences, 2019, 493: 57-74
[21]	Cai J F, Candès E, Shen Z W. A singular value thresholding algorithm for matrix completion[J]. SIAM Journal of Optimization, 2010, 20(4): 1956-1982
[22]	Boyd S, Neal P, Eric C. Distributed optimization and statistical learning via the alternating direction method of multipliers[J]. Foundations & Trends in Machine Learning, 2011, 3(1): 1-122
[23]	Kuang S F, Chao H Y, Li Q. Matrix completion with capped nuclear norm via majorized proximal minimization[J]. Neurocomputing, 2018, 316: 190-201
[24]	Fan J Q, Li R Z. Variable selection via nonconcave penalized likelihood and its oracle properties[J]. Journal of the American Statistical Association, 2001, 96(456): 1348-1360
[25]	Zhang T. Analysis of multi-stage convex relaxation for sparse regularization[J]. Journal of Machine Learning Research, 2010, 11: 1081-1107
[26]	Zou H, Li R Z. One-step sparse estimates in nonconcave penalized likelihood models[J]. Annals of Statistics, 2008, 36(4): 1509-1533
[27]	Garcia D. Robust smoothing of gridded data in one and higher dimensions with missing values[J]. Computational Statistics and Data Analysis, 2010, 54(4): 1167-1178