suhang9908 发表于 2009-2-23 17:06

与大师的近距离接触-听HHT创始人黄鄂授课有感

从不在BBS上发文章,因为觉得自己没有什么独特而有价值的东西与大家分享。我是个HHT的初学者,似乎只有两三个月,有点信号处理的常识,太多不懂的地方,所以从这个论坛上收获不少。今天之所以写,是因为前几天天听了HHT 之父——黄鄂的讲课,真的是感触颇深,一直拜读他的文章,没想到可以见到本人,还可以请教他一些问题,实在是幸运而不幸,不幸来源于自己实在研究的不够深入,以至于有了这样的机会,也没法问出什么尖锐的问题,有点不爽。黄鄂在我们学校一共讲了三天,由于某些内部的原因,我很晚才得到消息,只听了最后一天的课,不过就只这一天都让我受益匪浅,大师就是大师,现将大师的一些观念整理如下(很不好意思的加进了我自己的理解,望大家见谅):
1 黄老师对于HHT的热爱,或者说对于自己所从事的研究的热爱,是最让我影响深刻,也最值得我们铭记在心的,光凭这点,即便忽略掉他名字后面长长的title,都值得我们对他的尊重。
我一直不怎么相信名人简历里写的东西,国人忽悠的本事太利害了,可能是我经历得太多。俗话说,眼见为实,一点不假。黄老师说他总是在不停的思考,我信这不是刻意的强调,他在我们学校上课,从早9点讲到晚5点,中午休息3小时,这是我参加了的。晚上他还要思考我们学校项目的问题,我肯定他确实做到了,因为第二天,他就给出了建议,而且听者无不觉得有道理。所以这都是我所亲眼见到的实事,绝不是因为看了他的简历,就对他盲目的崇拜。
这位满头白发,但双眼散发着光芒的师者,是把自己提出的理论,把HHT当成自己的孩子一样,希望他能被大家接受,希望他能成才。黄老师说,每当他拿到一段数据,就想用HHT来分析,一直到分析出结果为止。他的这种不停滞的态度,真正的热情,太令我们感到惭愧了。我们以为从网上down几个程序,装模作样的改几行,画几个图,凑篇文章就以为自己在从事研究,再靠着这些东西去混硕士,混博士文凭。
我一开始并没有写专业相关的东西,而是提了些老生常谈,或许这才是决定人与人差距的地方。
2 黄老师对于Fourier 有着自己独到的见解。我们都知道,做信息处理,Fourier 是个万金油,至少很多年的实践经验和很多过去的学者这样认为,以致于我们拿到一段信号,第一反应就是做FT变换。这点没有错,错在Fourier也不是万能的,明明用Fourier没有得到有用的结果,为什么不去思考这种方法的弊端。黄老师认为很多以前的说法都是基于Fourier就是真理的前提,例如我们常看到关于instantaneous frequency 的种种悖论 ,(基本都是参考cohen的)都是基于这个假设的,如果Fourier在某些情况下不成立,这些悖论也就不是悖论了。Fourier是个global, linear的分析工具,这点黄老师一直强调,具体我就不写了,大家可以参考他的论文。
3 现在的市面上每年都有大量信息处理,分析的文章、理论、方法出现,深究其原因,是因为信号,图像的处理并不像物理,化学等学科,有自己的完整而严谨的系统,也就是说他们并没有一条固定的路可以走,我感觉随机性很大。例如图像分割,往往是用这种方法也行,用那种也行,或者是同一方法,用在这幅图效果很好,换一幅就不能用了。信号处理也有同样的问题。
4 基于3所提出的问题,我们做信号分析时要用到的一个强有力的工具就是我们的眼睛,这点黄老师在他的98年的文章中也有提到。好好的观察信号的原始图(一般是时间——幅度图),思考曲线为什么会这样走,再去选择分析的工具。
5 可能由于国内太急于求成的环境,很多人都是认死理。做信号处理的就只关心信号处理,全然不顾这是关于什么变化的信号,这个‘什么的’将是分析处理信号的突破口。打个比方,你想做旋转机械的故障检测,你拿到了采集到的转子的震动图,你就必须知道转子在正常的时候是什么样子,在发生故障时可能是什么样子,可能有哪几种故障,各是什么样的震动图。我相信这些知识的获得是可以请教搞机械的人,他们可能不懂信号处理,但没有这些常识,你就什么都做不了。
6 见多才能识广。这点和5点其实是一个意思。还是那句话,要多看,多问,多想。特别是要向自己的非同行请教。你可能没有见过高性能的战斗机分行,但你知道鸟在分行的时候做的是什么动作,其实都可以从中收到启发。黄老师来我们学校参观了桥梁健康检测系统,他说看了这个以后,再去看其他的桥就不会不知从何入手。
7 关于现有文献中提到的HHT没有一个rigorous mathematics foundation,这点在论坛中也常有人提到,黄老师给出了一句名言:不能因为我不知道身体的消化过程原理,就不吃饭了。(出自一位天才物理学家,因为黄老师用英文说的,我不记得名字了,似乎和Maxwell Equation 有关系)所以我们也不能因为不清楚HHT的 rigorous mathematics foundation,就不去用它。
在黄老师看来,这正是HHT的优势所在,HHT是一个超adaptive的方法,我们不清楚它到底是怎么回事,所以可以把它用在所有的信号上。也正因为它的adaptive的特性,我们在使用它的时候,也要灵活多变。其实这点也是基于3点的。打个比方:用EMD分解出了一系列分量,C1-C10,有时候要全部都用上,有时候其中一个分量的图比较接近我们的预想,这时就只用这个分量去画hilbert spectrum。这其中有很大一部分是来源于经验。所以还是那句,多做,做多了,就有feeling了。
黄老师说他本身数学不行,所以申请到了一笔经费,请到Princeton 应用数学系的Daubechies做带头人去寻找HHT的 rigorous mathematics foundation。Daubechies就是那个定义Daubechies正交小波的人,也是个牛人,没想到居然是位有点搞怪的美人。相信在这样一群人的带领下,找到HHT的 rigorous mathematics foundation指日可待。
8 最后,还有一点,搞研究,听大师布道,英文真的非常非常重要。黄老师给我们讲课,开始还注意尽量使用中文,随着讲课的深入,他情不自禁的不停飚English,希望English不太好的同学真的要恶补下才行。(同去的同学硬要补上的,我觉得如果一直都是看英文论文的人是不会有这方面的问题)

唠唠叨叨的写了一大堆,也不知道有没有传递出黄老师讲课的精髓,对大家有没有帮助。

吃书的老虎 发表于 2009-2-23 19:25

我也搞HHT,可惜没有你的运气。贵校是什么名字?

xiaoxiong702 发表于 2009-5-31 17:09

回复 楼主 suhang9908 的帖子

我使用了EMD一段时间处理效果有些还可以,但是进一步深入就不知该如何走了。看了楼主的帖子,也很惭愧,之前的效果也都是用别的东西得到的。下载了huang的授课视频,正准备好好研究,希望自己能努力发现适合处理我信号的方法。

[ 本帖最后由 xiaoxiong702 于 2009-5-31 17:10 编辑 ]

Robotech 发表于 2009-6-1 08:58

原帖由 suhang9908 于 2009-2-23 17:06 发表 http://www.chinavib.com/forum/images/common/back.gif
从不在BBS上发文章,因为觉得自己没有什么独特而有价值的东西与大家分享。我是个HHT的初学者,似乎只有两三个月,有点信号处理的常识,太多不懂的地方,所以从这个论坛上收获不少。今天之所以写,是因为前几天天听了 ...


如果Db女士能接受huang的这个任务,说不定还有戏。

想当初,刚看到EMD HHT的时候,就感觉到没有数学基础,所以自己很少信任EMD HHT

还曾跃跃欲试去建立一些数学基础,可惜作为一名工科小硕,虽然数模竞赛成绩挺好,但跟自己想象的水平相去还太远太远

希望Db女士能成功,给EMD-HHT一个正宗的出身

我们都知道,没有严谨的数学基础的信号处理方法,最终会被边缘化,被挤出历史舞台。

希望拯救了小波的Db女士,能拯救EMD HHT

Robotech 发表于 2009-6-1 09:40

还有补充一下,说出"Shall I refuse my dinner because I do not fully understand the process of digestion ?"这句话的,是Oliver Heaviside;此人是近代一个自学成才的工程学家、科学家,正规教育只有小学水平,但成果却十分丰硕;可惜,其人生比较悲惨,究其原因,就是他采用的研究方法不能与同时代的主流科学家们形成一致。

他创立的方法、得到的结论,跟当前的EMD HHT的处境有些类似之处:海维赛德被主流数学家瞧不起的原因之一就是,他的方法缺乏数学上
的严谨(faulted him for lack of rigor);而被实际工程人员批判的原因又是使用了过多的数学。

现转载B.P. Lathi的经典教材《Linear System and Signal Processing》中对海维赛德的介绍,供大家参考。

————————————————————————————————————————————————————————

textbook today mentions his name or contributions. With the passage of time,
Heaviside becomes more distant, although his studies had a major impact on many
aspects of modern electrical engineering. It was Heaviside who made transatlantic
communication possible by inventing cable loading, but no one ever mentions him
as a pioneer or an innovator in telephony. It was Heaviside who suggested the use
of inductive cable loading, but the credit is given to M. Pupin, who actually built
the first loading coil. In addition, Heaviside was;
• The first to find a solution to the distortionless transmission line;
• The innovator of lowpass filters;
• The first to write Maxwell's equations in modern form;
• The co-discoverer of rate energy transfer by an electromagnetic fieldj
• An early champion of the now-common phasor analysis;
• An important cont巾utor to the development of vec!,or analysis. In fact, he
essentially created the subject independently of Gibbs.
• An originator of the use of operational mathematics used to solve linear
integro-differential equations, which eventually led to rediscovery of the ignored
Laplace transform;
• The first to theorize (along with Kennelly of Harvard) that a conducting layer
(the Kennelly-Heaviside layer) of atmosphere exists, which allows radio waves
to follow earth's curvature instead of traveling off into space in a straight line;
• The first to posit that an electrical charge would increase in mass as its
velocity increases; an anticipation of an aspect of Einstein's special theory of
relativity.7 He also forecast the possibility of superconductivity.
Heaviside was a self-made, self-educated man with only an elementary school
education. who eventually became a pragmatically successful mathematical physi-
cist. He began his careerωa telegrapher, but increasing deafness forced him to


retire at the age of 24. He then devoted himself to the study of electricity. His
creative work was disdained by many professional mathematicians because of his
lack of formal education and his unorthodox methods
Heaviside's misfortune was that he was criticized both by mathematicians, who
faulted him for lack of rigor, and by men of practice, who faulted him for using too
much mathematics and thereby confusing students. Many mathematicians, trying
to find solutions to the distortionless transmission line, fai!ed because 110 rigorous
tools were available at the tirne. Heaviside succeeded because he used mathematics
not with rigor, but with illSight and illtuition. Using his much-m叫igned operational
method、Heaviside successfully attacked problems that the rigid mathematicians
could 1l0t solve: problems such as the flow of heat in a body of spatially varyillg
cOllductivity. Heaviside brilliantly used this method in 1895 to demollstrate a fatal
flaw in Lord Kelvin's determillation of the geological age of the earth by secular
coolingj he used the same flow of heat theory as for his cable analysis. Yet the
mathematicians of the Royal Society remained unmoved alld were not the least
impressed by the fact that Heaviside had found the answer to problems no one
else could solve. Mally mathematicians who examined his work dismissed it with
cOlltempt, asserting that his methods were either complete nonsellse or a rehash of
already-known ideas.5
Sir William Preece, the chief engilleer of the British Post Office, a savage critic
of Heaviside, ridiculed Heaviside's work as too theoretical and, therefore, leading to
faulty conclusions. Heaviside's work on transmission lines and loadillg was dismissed
by the British Post 0fice; this work might have remailled hidden, had not Lord
Kelvin himself publicly expressed admiration for it.5
Heaviside's operational calculus may be formally inaccurate, but in fact it allticipated
the currellt operational methods developed in more recent years.8 Although
his method was not fully understood, it provided correct results. When Heaviside
was attacked for the vague meanillg of his operational calculus, his pragmatic reply
was, "Shall I refuse my dinner because I do not fully understand the process of
digestion ?"
Heaviside lived as a bachelor hermit, often in near-squalid conditiollS, and died
largely unlloticed, ill poverty. His life demonstrates the arrogance and snobbish-
ness of the intelllectual establishment,which does not respect creativity unless it is presented in the strict language of the establishment.

[ 本帖最后由 Robotech 于 2009-6-1 09:48 编辑 ]

zhangnan3509 发表于 2009-12-29 20:21

希望大家不要多说这个问题了,咱们论坛是不提供共享服务的,大家把有限的资源和时间都用在实际问题的探讨上吧

cyp1967 发表于 2014-6-15 22:51

你们真幸运!{:{39}:}{:{39}:}{:{39}:}

hfutql 发表于 2014-7-17 09:59

{:{39}:}啊,强人!!!
页: [1]
查看完整版本: 与大师的近距离接触-听HHT创始人黄鄂授课有感