随着信息技术的持续进步，传感器技术已在各个领域得到了广泛的应用和发展。然而，面对复杂环境时，单一传感器往往无法满足高精度与高可靠性的需求。因此，多传感器数据处理技术逐渐成为研究焦点。近年来，虽然基于传感器的行为识别取得了很多的研究成果，但仍然存在诸多问题需要解决，主要包括如何有效地进行数据分割和特征提取，以及如何解决融合过程中的不确定性问题。本文主要工作内容如下：

传统的数据分割方法无法有效区分持续时间不同的人体活动信号，而现有方法通常依赖于特定活动信号的时域或频域特征，这导致了识别精度低、泛化能力差的问题。本文设计了一种基于置信度反馈机制的多传感器数据分割方法。该方法利用分类器的置信度作为反馈，并制定相应的调节机制实时修正滑动窗口的大小和重叠，使分类器能够适应并区分持续时间不同的人体行为。有效解决了某个序列的最佳参数设置可能在另一个分类器上无法实现正确分割的问题。在此基础上，对信号进行时域特征提取，并使用最小冗余最大相关（mRMR）方法进行特征筛选，从大量原始数据中提取出具有代表性的特征。最后，通过参数对比实验验证了该方法相较于传统方法在人体行为分类问题中具有更高的识别精度，达到95%，同时能够提升约30%的识别速度。

为了应对多传感器数据融合中存在的不确定性问题，本文设计了一种基于改进D-S证据理论的多传感器数据融合算法。首先采用三角模糊函数对证据理论的不确定性信息进行建模，生成基本概率分配（BPA）。在此基础上，设计了一种基于Jousselme距离和信息熵的多传感器数据融合方法。该方法通过Jousselme距离来衡量证据之间的冲突程度，得到证据可信度，同时采用信息熵来度量证据的模糊度；然后，对证据的BPA进行修正，并根据Dempster 组合规则对各证据进行融合。最后，利用区间概率转换进行决策。实验结果表明，该方法相较于传统方法与现有的改进方法精度分别提升了约3%和1.4%，为证据理论在多传感器行为识别中的应用提供了一种有效地理论依据及技术支撑。

With the continuous progress of information technology, sensor technology has been widely used and developed in various fields.  However, in the face of complex environments, a single sensor often cannot meet the needs of high precision and high reliability.  Therefore, multi-sensor data processing technology has gradually become the focus of research.  In recent years, although a lot of research results have been achieved in sensor-based behavior recognition, there are still many problems to be solved, including how to effectively segment data and extract features, and how to solve the uncertainty in the fusion process.  The main contents of this paper are as follows:

Traditional data segmentation methods can not distinguish human activity signals with different durations effectively, and existing methods usually rely on the time-domain or frequence-domain characteristics of specific activity signals, which leads to low recognition accuracy and poor generalization ability. In this paper, a multi-sensor data segmentation and feature extraction method based on confidence feedback mechanism is designed. In this method, the confidence of the classifier is used as feedback, and the corresponding adjustment mechanism is developed to correct the size and overlap of the sliding window in real time, so that the classifier can adapt to and distinguish the human behavior with different durations. It effectively solves the problem that the optimal parameter setting of one sequence may not achieve correct segmentation on another classifier. On this basis, the time domain features of the signal are extracted, and the minimum redundancy maximum correlation (mRMR) method is used for feature screening, and the representative features are extracted from a large number of original data. Finally, the experimental results show that the proposed method has a higher recognition accuracy of 95% and can improve the recognition speed by about 30% compared with traditional methods.

In order to deal with the uncertainty problem in multi-sensor data fusion, this paper designs a multi-sensor data fusion algorithm based on improved D-S evidence theory. Firstly, the uncertainty information of evidence theory is modeled by triangular fuzzy function, and the basic probability distribution (BPA) is generated. On this basis, a multi-sensor data fusion method based on Jousselme distance and information entropy is designed. Jousselme distance is used to measure the degree of conflict between the evidence and obtain the credibility of the evidence. Meanwhile, information entropy is used to measure the ambiguity of the evidence. Then, the BPA of the evidence was corrected and the evidence was fused according to Dempster's combination rule. Finally, the interval probability transformation is used to make the decision. The experimental results show that the accuracy of the proposed method is improved by about 3% and 1.4% respectively compared with the traditional method and the existing improved method, which provides an effective theoretical basis and technical support for the application of evidence theory in multi-sensor behavior recognition.

[1] Shaik T, Tao X, Higgins N, et al. Remote patient monitoring using artificial intelligence: Current state, applications, and challenges[J]. Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, 2023, 13(2): e1485.

[2] Manikandan N, Tadiboina S N, Khan M S, et al. Automation of Smart Home for the Wellbeing of Elders Using Empirical Big Data Analysis[C]//2023 3rd International Conference on Advance Computing and Innovative Technologies in Engineering (ICACITE). IEEE, 2023: 1164-1168.

[3] 罗国其.浅谈智能视频监控技术的发展趋势[J].网络安全和信息化,2023,(12):22-23.

[4] Geng X, Li Y, Sun Q. A Novel Short-Term Ship Motion Prediction Algorithm Based on EMD and Adaptive PSO–LSTM with the Sliding Window Approach[J]. Journal of Marine Science and Engineering, 2023, 11(3): 466.

[5] 钟伟华,黄达辉,王欣霖等.基于决策树模型的医院网络流量监测方法研究[J].微型电脑应用,2023,39(04):91-93+97.

[6] Li J H, Tian L, Wang H, et al. Segmentation and recognition of basic and transitional activities for continuous physical human activity[J]. IEEE access, 2019, 7: 42565-42576.

[7] Ma C, Li W, Cao J, et al. Adaptive sliding window based activity recognition for assisted livings[J]. Information Fusion, 2020, 53: 55-65.

[8] Alhammad N, Al-Dossari H. Dynamic segmentation for physical activity recognition using a single wearable sensor[J]. Applied Sciences, 2021, 11(6): 2633.

[9] 李华昊,郇战,耿宏杨,等.基于时频域特征的人员连续动作分段识别[J].重庆邮电大学学报(自然科学版),2021,33(05):877-884.

[10] Wan J, O’grady M J, O’Hare G M P. Dynamic sensor event segmentation for real-time activity recognition in a smart home context[J]. Personal and Ubiquitous Computing, 2015, 19: 287-301.

[11] Chen D, Yongchareon S, Lai E M K, et al. Hybrid fuzzy c-means CPD-based segmentation for improving sensor-based multiresident activity recognition[J]. IEEE Internet of Things Journal, 2021, 8(14): 11193-11207.

[12] 敬冉,李遥,秦朝倩.物联网领域多传感器数据融合技术的应用[J].智慧中国,2023,(10):85-86.

[13] Ounoughi C, Yahia S B. Data fusion for ITS: A systematic literature review[J]. Information Fusion, 2023, 89: 267-291.

[14] Shafer G. A mathematical theory of evidence[M]. Princeton university press, 1976.

[15] Dempster A P. Upper and lower probabilities induced by a multivalued mapping[M]// Classic works of the Dempster-Shafer theory of belief functions. Berlin, Heidelberg: Springer Berlin Heidelberg, 2008: 57-72.

[16] Qiang C, Deng Y. A new correlation coefficient of mass function in evidence theory and its application in fault diagnosis[J]. Applied Intelligence, 2022, 52(7): 7832-7842.

[17] Wang Y C, Wang J, Huang M J, et al. An evidence combination rule based on a new weight assignment scheme[J]. Soft Computing, 2022, 26(15): 7123-7137.

[18] Long H, Peng Z, Deng Y. A new structure of the focal element in object recognition[J]. International Journal of Intelligent Systems, 2022, 37(2): 1430-1453.

[19] Yang S, Huang J, Qin Y, et al. Evaluation of Concrete Durability Based on Cloud Model and DS Evidence Theory[C]//Journal of Physics: Conference Series. IOP Publishing, 2022, 2148(1): 012054.

[20] Yager, R.R. On the Dempster-Shafer framework and new combination rules. Inf. Sci. 1987, 41, 93–137.

[21] Sun Y, Li C, Li G, et al. Gesture recognition based on kinect and sEMG signal fusion[J]. Mobile Networks and Applications, 2018, 23: 797-805.

[22] Yang K, Tan T, Zhang W. An Evidence Combination Method based on DBSCAN Clustering[J]. Computers, Materials & Continua, 2018, 57(2).

[23] Si L, Wang Z B, Jiang G. Fusion recognition of shearer coal-rock cutting state based on improved RBF neural network and DS evidence theory[J]. IEEE Access, 2019, 7: 122106-122121.

[24] Höhle U. Entropy with respect to plausibility measures[C]//Proc. of 12th IEEE Int. Symp. on Multiple Valued Logic, Paris, 1982. 1982.

[25] Deng Y. Deng entropy[J]. Chaos, Solitons & Fractals, 2016, 91: 549-553.

[26] Yager R R. Entropy and specificity in a mathematical theory of evidence[J]. Classic works of the Dempster-Shafer theory of belief functions, 2008: 291-310.

[27] Jousselme A L, Liu C, Grenier D, et al. Measuring ambiguity in the evidence theory[J]. IEEE Transactions on Systems, Man, and Cybernetics-Part A: Systems and Humans, 2006, 36(5): 890-903.

[28] Zheng H, Tang Y. Deng entropy weighted risk priority number model for failure mode and effects analysis[J]. Entropy, 2020, 22(3): 280.

[29] Yan H, Deng Y. An improved belief entropy in evidence theory[J]. IEEE Access, 2020, 8: 57505-57516.

[30] Gou L, Zhang J, Li N, et al. Weighted assignment fusion algorithm of evidence conflict based on Euclidean distance and weighting strategy, and application in the wind turbine system[J]. PLoS One, 2022, 17(1): e0262883.

[31] Hua Y, Sun Y, Xu G, et al. A fault diagnostic method for oil-immersed transformer based on multiple probabilistic output algorithms and improved DS evidence theory[J]. International Journal of Electrical Power & Energy Systems, 2022, 137: 107828..

[32] Jing M, Tang Y. A new base basic probability assignment approach for conflict data fusion in the evidence theory[J]. Applied Intelligence, 2021, 51: 1056-1068.

[33] Reiss A, Stricker D. Introducing a new benchmarked dataset for activity monitoring[C]//2012 16th international symposium on wearable computers. IEEE, 2012: 108-109.

[34] Choudhury N A, Soni B. In-depth analysis of design & development for sensor-based human activity recognition system[J]. Multimedia Tools and Applications, 2023: 1-40.

[35] 曲梦杰.基于可穿戴传感的跨用户行为识别技术研究[D].浙江大学, 2023.000195

[36] Mirzaei A, Bagheri H, Sattari M. Data level and decision level fusion of satellite multi-sensor AOD retrievals for improving PM2. 5 estimations, a study on Tehran[J]. Earth Science Informatics, 2023, 16(1): 753-771.

[37] 钱倍奇,陈谦,张政伟等.基于异构数据特征级融合的多任务暂态稳定评估[J].电力系统自动化,2023,47(09):118-128.

[38] 南鹏,群诺,温瑶等.基于决策级融合策略的中文网络模因图片判别方法研究[J].中央民族大学学报(自然科学版),2023,32(02):24-30.

[39] Dehghani A, Sarbishei O, Glatard T, et al. A quantitative comparison of overlap** and non-overlap** sliding windows for human activity recognition using inertial sensors[J]. Sensors, 2019, 19(22): 5026.

[40] Peng H, Long F, Ding C. Feature selection based on mutual information criteria of max-dependency, max-relevance, and min-redundancy[J]. IEEE Transactions on pattern analysis and machine intelligence, 2005, 27(8): 1226-1238.

[41] Ma L, Yao W, Dai X, et al. A New Evidence Weight Combination and Probability Allocation Method in Multi-Sensor Data Fusion[J]. Sensors, 2023, 23(2): 722.

[42] 王谦,高海波,赵云瑞等.基于DS证据理论的极地邮轮减摇鳍选型评价[J].舰船科学技术,2023,45(05):54-60.

[43] 付威,王欣.一种新的GBPA生成方法及其在模式识别中的应用[J].控制与决策,2024,39(03):994-1002.

[44] 吕天宝.基于信息融合的软件项目风险分析[D].重庆大学.2022.003844.

[45] Fu C, Zhan Q, Liu W. Evidential reasoning based ensemble classifier for uncertain imbalanced data[J]. Information Sciences, 2021, 578: 378-400.

[46] Jousselme A L, Grenier D, Bossé É. A new distance between two bodies of evidence[J]. Information fusion, 2001, 2(2): 91-101.

[47] Ren Z, Liao H. Combining conflicting evidence by constructing evidence’s angle-distance ordered weighted averaging pairs[J]. International Journal of Fuzzy Systems, 2021, 23: 494-505.

[48] Staerman G, Laforgue P, Mozharovskyi P, et al. When ot meets mom: Robust estimation of wasserstein distance[C]//International Conference on Artificial Intelligence and Statistics. PMLR, 2021: 136-144.

[49] Piccoli B, Rossi F. Generalized Wasserstein distance and its application to transport equations with source[J]. Archive for Rational Mechanics and Analysis, 2014, 211: 335-358.

[50] Vallender S S. Calculation of the Wasserstein distance between probability distributions on the line[J]. Theory of Probability & Its Applications, 1974, 18(4): 784-786.

[51] Panaretos V M, Zemel Y. Statistical aspects of Wasserstein distances[J]. Annual review of statistics and its application, 2019, 6: 405-431.