基于强化学习的多模态场景人体危险行为识别方法

doi:10.3969/j.issn.0255-8297.2021.04.008

摘要/Abstract

摘要： 在多模态场景下，常规人体危险行为识别方法对人体危险行为的识别精度较低，于是提出了基于强化学习的多模态场景人体危险行为识别方法。首先根据强化学习的特征提取算法获取多模态场景人体危险行为特征集，其次基于强化学习数据决策提取多模态场景人体危险行为，构建人体危险行为模糊识别模型。最后将上述人体危险行为特征子集代入模型，计算不同感官下危险行为的隶属度，实现多模态场景人体危险行为的识别。实验结果表明：该方法对危险行为的识别准确率较高，其识别延迟时间低于300 ms。

关键词: 强化学习, 多模态, 场景, 行为识别

Abstract: In multimodal scenes, conventional human dangerous behavior recognition methods generally perform low recognition accuracy. Therefore, this paper proposes a human dangerous behavior recognition method based on reinforcement learning. Firstly, a feature extraction algorithm of reinforcement learning is used to obtain feature subsets of human dangerous behavior in multimodal scenes. Secondly, human dangerous behaviors in multimodal scenes are extracted by reinforcement learning data decision-making, and a fuzzy recognition model of human dangerous behavior is constructed. Finally, by bringing the obtained feature subsets of human dangerous behavior into the model and calculating the membership degree of dangerous behavior under different senses, the recognition of human dangerous behavior in multimodal scenes can be realized. Experimental results show that the method in this paper has a high recognition accuracy and a recognition delay of less than 300 ms.

Key words: reinforcement learning, multimodality, scene, behavior recognition

中图分类号:

TP391

张晓龙, 王庆伟, 李尚滨. 基于强化学习的多模态场景人体危险行为识别方法[J]. 应用科学学报, 2021, 39(4): 605-614.

ZHANG Xiaolong, WANG Qingwei, LI Shangbin. Recognition Method of Human Dangerous Behavior in Multimodal Scenes Using Reinforcement Learning[J]. Journal of Applied Sciences, 2021, 39(4): 605-614.

参考文献

[1] 唐超, 王文剑, 张琛, 等. 基于RGB-D图像特征的人体行为识别[J]. 模式识别与人工智能, 2019, 32(10):901-908.Tang C, Wang W J, Zhang C, et al. Human behavior recognition based on RGB-D image features[J]. Pattern Recognition and Artificial Intelligence, 2019, 32(10):901-908. (in Chinese)
[2] 陈煜平, 邱卫根. 基于视觉的人体行为识别算法研究综述[J]. 计算机应用研究, 2019, 36(7):1927-1934. Chen Y P, Qiu W G. Overview of human behavior recognition algorithm based on vision[J]. Computer Application Research, 2019, 36(7):1927-1934. (in Chinese)
[3] 鹿天然, 于凤芹, 陈莹. 一种基于线性序列差异分析降维的人体行为识别方法[J]. 计算机工程, 2019, 45(3):237-241, 249. Lu T R, Yu F Q, Chen Y. A method of human behavior recognition based on dimension reduction of linear sequence difference analysis[J]. Computer Engineering, 2019, 45(3):237-241, 249. (in Chinese)
[4] 叶丹, 李智, 王勇军. 基于SPLDA降维和XGBoost分类器的行为识别方法研究[J]. 微电子学与计算机, 2019, 36(6):35-39. Ye D, Li Z, Wang Y J. Research on behavior recognition method based on SPLDA dimensionality reduction and xgboost classifier[J]. Microelectronics and Computer, 2019, 36(6):35-39. (in Chinese)
[5] 王萍, 庞文浩. 基于视频分段的空时双通道卷积神经网络的行为识别[J]. 计算机应用, 2019, 39(7):2081-2086. Wang P, Pang W H. Behavior recognition of space time dual channel convolutional neural network based on video segmentation[J]. Computer Applications, 2019, 39(7):2081-2086. (in Chinese)
[6] 袁亚军, 李菲菲, 陈虬. 基于复合特征及深度学习的人群行为识别算法[J]. 计算机科学, 2019, 46(6):305-310. Yuan Y J, Li F F, Chen Q. Crowd behavior recognition algorithm based on composite features and deep learning[J]. Computer Science, 2019, 46(6):305-310. (in Chinese)
[7] 佟瑞鹏, 张艳伟. 人工智能技术在矿工不安全行为识别中的融合应用[J]. 中国安全科学学报, 2019, 29(1):7-12. Tong R P, Zhang Y W. Fusion application of artificial intelligence technology in miners' unsafe behavior recognition[J]. Chinese Journal of Safety Science, 2019, 29(1):7-12. (in Chinese)
[8] 郑兴华, 孙喜庆, 吕嘉欣, 等. 基于深度学习和智能规划的行为识别[J]. 电子学报, 2019, 47(8):1661-1668. Zheng X H, Sun X Q, Lu J X, et al. Behavior recognition based on deep learning and intelligent planning[J]. Acta Electronica Sinica, 2019, 47(8):1661-1668. (in Chinese)
[9] 罗会兰, 王婵娟. 行为识别中一种基于融合特征的改进VLAD编码方法[J]. 电子学报, 2019, 47(1):49-58. Luo H L, Wang C J. An improved VLAD coding method based on fusion features in behavior recognition[J]. Chinese Journal of Electronics, 2019, 47(1):49-58. (in Chinese)
[10] 王莉, 张紫烨, 牛群峰, 等. 基于MPU9250和MS5611的人体姿态检测系统设计[J]. 电子器件, 2019, 42(4):978-983. Wang L, Zhang Z Y, Niu Q F, et al. Design of human posture detection system based on MPU9250 and MS5611[J]. Electronic Devices, 2019, 42(4):978-983. (in Chinese)
[11] Dai C, Liu X, Lai J, et al. Human behavior deep recognition architecture for smart city applications in the 5G environment[J]. IEEE Network, 2019, 33(5):206-211.
[12] 殷晓玲, 夏启寿, 陈晓江, 等. 基于智能手机感知的人体运动状态深度识别[J]. 北京邮电大学学报, 2019, 42(3):43-50. Yin X L, Xia Q S, Chen X J, et al. Deep recognition of human motion state based on smart phone perception[J]. Journal of Beijing University of Posts and telecommunications, 2019, 42(3):43-50. (in Chinese)
[13] 杨世强, 罗晓宇, 李小莉, 等. 基于DBN-HMM的人体动作识别[J]. 计算机工程与应用, 2019, 55(15):169-176. Yang S Q, Luo X Y, Li X L, et al. Human motion recognition based on DBN-HMM[J]. Computer Engineering and Applications, 2019, 55(15):169-176. (in Chinese)
[14] Kaur M, Kaur G, Sharma P K, et al. Binary cuckoo search metaheuristic-based supercomputing framework for human behavior analysis in smart home[J]. The Journal of Supercomputing, 2020, 76(4):2479-2502
[15] 刘建伟, 高峰, 罗雄麟. 基于值函数和策略梯度的深度强化学习综述[J]. 计算机学报, 2019, 42(6):1406-1438. Liu J W, Gao F, Luo X L. A review of deep reinforcement learning based on value function and strategy gradient[J]. Acta Computer Sinica, 2019, 42(6):1406-1438. (in Chinese)
[16] Wang Z, Guo B, Yu Z, et al. Wi-Fi CSI-based behavior recognition:from signals and actions to activities[J]. IEEE Communications Magazine, 2018, 56(5):109-115.
[17] Ding W, Liu K, Belyaev E, et al. Tensor-based linear dynamical systems for action recognition from 3D skeletons[J]. Pattern Recognition, 2018, 77:75-86.
[18] Dapogny A, Bailly K, Dubuisson S. Confidence-weighted local expression predictions for occlusion handling in expression recognition and action unit detection[J]. International Journal of Computer Vision, 2018, 126(2/3/4):255-271.
[19] Pino M, Montaño S, Agudelo K, et al. Emotion recognition in young male offenders and non-offenders[J]. Physiology & Behavior, 2019, 207:73-75.
[20] Kell A J E, Yamins D L K, Shook E N, et al. A task-optimized neural network replicates human auditory behavior, predicts brain responses, and reveals a cortical processing hierarchy[J]. Neuron, 2018, 98(3):630-644
[21] Weng J, Jiang X, Zheng W L, et al. Early action recognition with category exclusion using policy-based reinforcement learning[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2020, 30(12):4626-4638.
[22] Lu C, Hu F, Cao D, et al. Virtual-to-real knowledge transfer for driving behavior recognition:framework and a case study[J]. IEEE Transactions on Vehicular Technology, 2019, 68(7):6391-6402.
[23] Jiang Z, Crookes D, Green B D, et al. Context-aware mouse behavior recognition using hidden Markov models[J]. IEEE Transactions on Image Processing, 2018, 28(3):1133-1148.