法国国家科学院院长 Patrick Flandrin 院士给我院师生做学术报告

摘要：Dealing with data in a well-adapted representation space is a key for understanding, modifying or compressing observed signals. In nonstationary situations where spectral features are considered as varying over time, combining the—usual but mutually exclusive—time and frequency variables offers such a natural framework which formalizes the intuitive idea of a musical score. Sounds are transformed this way as readable images in a joint time-frequency plane and, conversely, time-frequency patterns can be turned into audible waveforms. As it is the rule in signal processing, with the Fourier analysis as the most prominent example in a stationary setting, the ‘art of time-frequency analysis’ is to design and use meaningful, well-defined and efficient tools which lie at the intersection of physics, applied mathematics and computer science, even when stationarity does not hold anymore. In this lecture, basics of classical time-frequency analyses will be briefly recalled, and recent advances related respectively to varying frequencies, localized oscillations and possibly non-harmonic modes will be discussed more specifically, as three ways of going beyond the classical Fourier analysis whose publication is just celebrating this year its 200th anniversary. 

报告人简历：Patrick Flandrin教授，法兰西科学院院士、法国里昂高等师范学院教授、法国科学研究中心主任研究员。现任法兰西科学院院长、IEEE 会士、欧洲信号处理学会会士。2009至2019年，担任法国信号和图像处理协会主席。长期致力于信息与信号处理方面研究工作，在非平稳信号时频分析、多分辨率小波分析等领域取得了卓越成就和重要贡献。曾获国际光学工程学会“小波先驱奖”、IEEE信号处理学会“技术成就奖”、 法国国家科学研究中心银奖、法国荣誉军团骑士勋章等重要奖项。发表相关学术论文300余篇，并撰写了多本专著，如《时频分析探索》(剑桥大学出版社，2018年)。