無線感測網路常用於環境變數的收集與估測狀態，像是溫度、壓力、亮度等等，過去的文獻中已探討使用有限個數的感測器節點去估計，並且在有最佳佈署位置的感測器節點中負責回報該位置的測量值給融合中心(Fusion center, FC)，最後再估計整個高斯隨機場。但如果要再加強對整個隨機場的估測效能，則必須增加感測器數量、傳輸頻寬和總體能源的消耗。本論文中，我們在不增加感測器數量及傳輸頻寬的情況下，採用離散量化的估測方法使得系統消耗功率能有效地下降，我們也使用一種讓多位元平均變成二位元來隨機傳送的方法，以達到傳送能量減少的目的。另外為了使估計分析的環境更接近實際情況並且能夠有效提升估測效能，我們考慮前後時間互有相關聯的高斯隨機場，而因為一般的估計方式是直接傳輸原始的資料，為了降低所佔用的通道頻寬成本，我們使用新的過濾數據方法，減少傳送相似的資料來達到節省能源的目的，最後由模擬結果分析來驗證我們提出的方法。

Wireless sensor network is usually used for collecting environmental measurements, such as temperature, pressure and brightness, and inferring the environmental status based on the collected measurements. The problem of the construction of a Gaussian random field using a finite number of sensor nodes has been considered in the literature. After deploying the sensor nodes to the best positions, sensor nodes report their observation taken at locations to the fusion center. However, to improve the reconstruction performance, it is necessary to increase the number of sensor nodes, transmission bandwidth and the power consumption. This thesis, in the situation without extra sensors, we use a quantization method to reduce transmission bandwidth and make the power consumption more effectively. Transmitting average binary bits randomly to the fusion center is also proposed. In order to make the considered system scenario close to real environment, we consider the Gaussian random field with temporal correlation. The censoring sensors approach, in which the close value of data are transmitted once instead of being transmitted at each time slot, is employed for energy saving. The results show that the proposed scheme can achieve a lower mean square error and reduce the power consumption.

摘要i
Abstract ii
1 序論1
1.1 前言 1
1.2 研究方法 2
1.3 論文貢獻 3
1.4 論文架構 3
2 重建具有時間關聯性的隨機場 4
2.1 隨機場重建問題 4
2.2 時間序列相關性模型 7
2.3 過濾數據 8
2.4 過濾方式 10
2.5 模擬結果 11
3 以量化訊息重建隨機場 14
3.1 量化估計模型 14
3.1.1 量化誤差 15
3.1.2 量化重建值 16
3.1.3 量化邊界值 17
3.2 量化程序 18
3.2.1 改變量化區間數 21
3.3 以不同量化區間探討平均位元資料的傳送 23
3.4 模擬結果 25
4 結論 27
參考文獻 28

[1] Y. Yang, and R. S. Blum,“Sensor Placement in Gaussian Random Field Via Discrete
Simulation Optimization,”IEEE Signal Processing Letters, vol. 15, 2008.
[2] I. F. Akyildiz, M. C. Vuran, and O. B. Akan, “On Exploiting Spatial and Temporal Cor-
relation in Wireless Sensor Networks,”School of Electrical Computer Engineering Georgia
Institute of Technology, Atlanta, June 2004.
[3] S. P. Lloyd, “Least Squares Quantization in PCM,”IEEE Transactions on Information
Theory, vol. IT-28, no. 2, March 1982.
[4] S. Seo, M. Wallat, T. Graepel, and K. Obermayer, “Gaussian process regression: active
data selection and test point rejection,” in Proc. Int. Joint Conf. Neural Networks, pp.
241-246, July 2000.
[5] C. Guestrin, A. Krause, and A. P. Singh, “Near-optimal sensor placements in Gaussian
processes,” in Proc. Int. Conf. Machine Learning, pp. 265-272, August 2005.
[6] E. J. Msechu, and G. B. Giannakis, “Sensor-Centric Data Reduction for Estimation With
WSNs via Censoring and Quantization,”IEEE Transactions on Signal Processing, vol. 60,
no. 1, January 2012.
[7] H. Rue and L. Held, Gaussian Markov Random Fields: Theory and Applications, Monographs
on Statistics and Applied Probability, vol. 104, Chapman and Hall, London, 2005.
[8] P. F. Panter andW. Dite, “Quantization distortion in pulse-count modulation with nonuni-
form spacing of levels,”Proc. I. R. E., vol. 39, pp. 44-48, 1951.
[9] N. A. C. Cressie, Statistics for Spatial Data, Revised Edition. New York: Wiley-
Interscience, September 1993.
[10] J. J. Zhao, Z. C. Wei, Z. H. Li, H. Liu, and B. Lian, “Research of Multi-type Data Fusion
in Sensor Networks,”Applocation Research of Computers, vol. 29, no. 8, August 2012.
[11] M. Mayer, “Quantization of Images and Lloyd’s Algorithm,”Institute of Communications
and Radio-Frequency Engineering, Vienna University of Technology, September 2010.
[12] A. Deligiannakis, and Y. Kotidis, “Exploiting Spatio-temporal Correlations for Data
Processing in Sensor Networks,”Geosensor Networks, Second International Conference,
Boston, USA, pp. 45-65, 2008.