主题：深度学习与复调声音事件检测

时间、地点：2018年8月30日院楼106

报告人:Ruili Wang，新西兰梅西大学人工智能专业教授

报告内容：

The aim of automatic sound event detection is to recognize sound events which present in a continuous acoustic signal. Polyphonic sound detection tackles the situations where multiple sound events happen simultaneously. Recently, polyphonic sound event detection has been utilized in a variety of applications, including scene recognition for mobile robots, monitoring in healthcare, surveillance in living environments, and video captioning.  Polyphonic sound event detection can also be used in detecting the songs and calls of birds or whales. In our research, two challenges in polyphonic sound detection are identified and explored. After that, two approaches proposed are to address the challenges. The overall aim of this research is to develop effective and efficient approaches for polyphonic sound event detection.

报告人介绍：2003年获得爱尔兰都柏林城市大学计算机科学专业博士学位。其研究领域涉及语音与语言处理、机器学习和数据挖掘、计算机视觉与图像处理等方面，发表学术论文共120余篇，其中68篇为期刊论文。目前担任IEEE Transactions on Emerging Topics in Computational Intelligence (IEEE)、Knowledge and Intelligent Systems (Springer)、Applied Soft Computing (Elsevier)以及Neurocomputing (Elsevier)等国际期刊的副编辑(或编委会成员)。指导了17名博士和8名硕士完成学业。在基于机器学习的语音处理方面，于2013年获得了新西兰Marsden项目基金资助，并于2017年获得了新西兰国家科学挑战的种子项目基金资助。