IRIS Facebook粉專

廖弘源 (IEEE Fellow)

中央研究院資訊科學研究所 所長




東元獎, 台灣 (2016)

台灣人工智慧學會博士論文指導獎, 台灣 (2014)

中華民國資訊學會博士論文指導獎, 臺灣 (2014)

一零二年度國科會傑出研究獎, 臺灣 (2013–2016)

中華民國影像學會最佳博士論文指導獎, 台灣 (2013)

IEEE Fellow, 美國 (2013–present)

ACM Multimedia Conference Doctoral Symposium Best Paper Award(與陳殷盈及徐宏民), 日本奈良 (2012)

中央研究院深耕計畫獎, 台灣 (2010–2014)

九十九年度國科會傑出研究獎, 台灣 (2010–2013)


熊博安 (IET Fellow)

國立中正大學資訊工程學系暨研究所 教授

Landslides could cause huge threats to lives and property damages. In a landslide prediction system, environmental information can be collected through sensors to detect the possibility of landslide occurrences. However, the data collected by wireless sensor network systems (WSNs) may be lost due to sensor failures, external interferences or other environmental factors, which may in turn affect the accuracy of landslide predictions.

To address the issue of missing data, this talk proposes a novel data reconstruction method using spatio-temporal relationships among heterogenous data including rainfall intensity, soil moisture, slope, and slope direction. For spatial considerations, in addition to considering the distance between sensor nodes, the terrain of the area (such as slope and slope direction) and the area of rainfall are also considered. For temporal considerations, the data trend in the past period of time is leveraged along with the correlation between heterogeneous data (such as rainfall and soil moisture). A Convolutional Long Short-Term Memory (ConvLSTM) deep neural network model is proposed to predict missing data which are then used to compensate for the missing data. Compared to other reconstruction methods, our proposed method achieves significantly better accuracy. Irrespective of the pattern of how data is missing (random missing, gradual fading, or spatial event), and even if almost all 90% of data are lost, the RMSE value for the error in the SHALSTAB Factor of Safety (FS) in our proposed method is less than 0.3, whereas that in the conventional LSTM model, it is as high as 4.9 for spatial event, 3.69 for random missing, and 1.76 for gradual fading. Note that FS < 1 is the condition for detecting landslides and thus an error greater than 1 is unacceptable for any kind of detection or prediction.

Note that the deep neural-network based model for data reconstruction can be applied to any kind of wireless sensor network data loss, irrespective of the application domain.

Pao-Ann Hsiung, Ph.D., received B.S. in Mathematics and Ph.D. in Electrical Engineering from the National Taiwan University, Taiwan, in 1991 and 1996, respectively. He is currently a full professor in the Department of Computer Science and Information Engineering, the Director of Smart Living Technology Research Center, National Chung Cheng University (CCU), Taiwan. He is currently also the Vice-Director of the Taiwan-India Joint Research Center on Artificial Intelligence, a joint effort between the Indian Institute of Technology (IIT) Ropar, India and CCU. He was the Department Chair from August 2011 to January 2016 and the Dean of International Affairs from February 2016 to July 2017. He was also the Director General of the Intelligence Technology Department, Chiayi City Government in 2018 (嘉義市政府智慧科技處處長). He has published more than 285 papers in international journals and conferences. He was the recipient of the 2010 CCU Outstanding Research Award, the 2001 ACM Taipei Chapter Kuo-Ting Li Young Researcher, and the 2004 CCU Young Scholar Research Award. Dr. Hsiung is an IET fellow, a senior member of the IEEE and the ACM, and a life member of the IICM. His main research interests include innovative teaching pedagogy, cyber-physical systems, reconfigurable computing, smart traffic optimization, and smart grid design. He can be contacted at More details can be found at