Mouth detection for yawn monitoring......

I need a code for mouth detection.................. specially i need to detect yawn for my project work..please anyone help.

[Jan]

@Image Analyst: Please tell me, that there are some projects in the field of image analysis, which are really useful for the mankind. If not I'm going to ask TMW to drop the Image Processing Toolbox support for Matlab.

I do not mean, that working on non-serious projects is a waste of time, or even if it is, wasting time is bad or evil. Any kind of art is "non-serious" in this sense, because any person could live without it. But without doubt the human culture could not exist without art, poetry, musik and fine arts. So this is not a criticism. I personally have implemented a tool, which correlates the moon-phase and the zodiac sign of a set of IR-cameras with the daily changing quality of their output. So you see, I have no problem with toy-applications. I'm only resignated, because I've seen too many questions in this forum about projects, which could be solved by very young children after one simple sentence for the instructions.

I know that automatic cancer recognition is not a toy-application. But on the other hand I'd never trust a diagnosis, which is based on an algorithm for automatic image registration, even if the author have asked in this forum for assistence (and this has happend several times). I'm sure that you, Image Analyst, and other experienced scientists working in this field, know serious and useful applications apart from the number plate and face recognition systems which are able to implement the horro scenarios from the science fiction literature of the 1940'th and 1980'th. I'd appreciate some notes, which would e.g. encourage professors to give more challenging final year projects.

[Image Analyst]

I see questions that could be useful real world applications, and I see others that are clearly just some exercise for a classroom, like identifying simple geometric shapes, coins, or fruit from a variety. I have no problem with classroom assignments because that's how they learn to do the more difficult applications, like medical image analysis, surveillance, OCR, CBIR, etc. But I wish they'd tell us what the project is for and the expected time frame, like it's a week long home work assignment, or a semester long senior project, or something for their masters or Ph.D. I sometimes see things that look too complicated for what I think the student was assigned and feel bad for them, like the recent mango detection problem. I think that the projects should be matched to the students' abilities. I'm surprised at the one fellow here who is supposedly an instructor for a course on MATLAB image processing and he constantly asks questions on how to do certain kinds of image processing, many of which are fairly easy and should certainly be easy for him since he's the instructor. So why is he asking? Asking how to do something that he will assign to students even he himself does not know how to do it. Odd. If he can't do it, why does he expect his students to be able to do it?

There are definitely real world projects that are worth real world money. The projects I work on affect literally affect almost everyone on planet. I'd say probably about 4 billion people have personally touched and used a product that I have worked on, and that's not an exaggeration. Sometimes I see really high tech applications that are really cool but are only used by a handful of people, like a few dozen or hundreds, while I work on things that a few billion people have used and benefited from. But they all have their place.

When I read this, I noted the lack of context, which is all too common. I assumed that this was for something like drowsy driver detection, but I could be wrong. And that is a fine application which could save lives. But I don't know if it's a student project where the professor supplied the student with some mouth shut videos and others with obvious wide yawns. If so, it could be rather easy. On the other hand if you need to detect heads that are turning, mouths that are open due to talking, or (worse) singing, then you can see how it could become very complicated and uncertain in a hurry. I'm also surprised at the number of people who ask for advice on image processing and think that they can get good advice without sharing an image. Plus the people who ask for some esoteric application as "part of their research" as if it's off the shelf software, for example (made up) "I really need software to recognize Dalmatians for my dog spot counting thesis. Please send it to me." Also a concern is the people who say "How do I do X?" and "X" is very specific because they think they need to do that to accomplish "Y". But they don't tell you what "Y" is! If they did tell you "Y", you could tell them that going after X is a complete waste of time and you really need to do "Z". Again, giving the context, the bigger picture, is always helpful.

Oh well, got to get back to my real work...

[Walter Roberson]

Maybe this is a project to go along with the automated classroom attendance projects, with this one going the step further to determine whether the student is paying attention...

[Image Analyst]

Professors care whether students are paying attention?

[Walter Roberson]

If public school teachers are going to be paid according to their "effectiveness", it follows that it is necessary to measure whether the students are paying attention.

[Image Analyst]

True. And if you wanted to make it a lot easier, you'd use a thermal infrared camera instead of a visible light camera, because the mouth is so much hotter than the face or clothing. See the mouth on the girl at http://www.healthjockey.com/2009/09/01/infrared-technology-may-prove-beneficial-for-motor-impaired-people/

[Sven]

And soon after implementation professors will start buying students ice cream to munch on in class... everyone wins.

[Jan]

Yawning is one of the unsolved problems in biology, physiology and epistemology: The theory that human are yawning when they have a deficite of oxygene can easily rejected. 1. A deficite of oxygen causes unconsciousness. 2. Babies inside the mothers womb yawn also even in the absence of air. 3. Fish yawn also.

Obviously the main reason for yawning is an interchange of information, because yawning causes other individual to yawn also (even if they belong to another species, but less effective then). Therefore it could be a serious application to let technicel instruments detect the user's yawning in realtime to allow a reaction with a high usability level: The robot, the software (Matlab 9.0) or the car have to yawn also to increase the level of harmony.

One of my fellow students in physics have applied tools for image analysis for a frequently used product also: He determined the perforation crack elongation resistance of different toilet papers automatically. "Perforationsbruchdehnungswiderstand", I love this term. This was needed to reduce the danger of plugging in the sewerage. But I'm not sure, if up to 4 billion people have access to a sewerage at all. Therefore it seems, like there is an even more fundamental and common object used almost all over the world. Enigmatic. Perhaps a toothbrush?

[Sven]

Yeah IA, colour me intrigued too. If we're talking "touched" an object (that requires image processing) then I'd guess laptop trackpad but I'd have to go with units sold rather than individuals used to reach 4 billion ... other options may be:

• Digital camera IP algorithm
• Some kind of generic scantron (school tests/government forms) patent that's now used all over
• Automatic sliding doors
• secret government fingerprinting via toothbrushes (Jan, that seems the obvious way to get IP into toothbrushes)
• IP that sorted out soda cans during manufacturing

... are any of these close?

[Image Analyst]

Let's just say that I work on an incredibly wide variety of products and that our company's products are available to 95% of the worlds consumers. And over the past 25 years I've worked on most of our products. I'm pretty certain you have (own) some of our products now, and you'd definitely, without a doubt, you know of several of them. Most people would be totally amazed and astounded if they knew the amount and kind of high tech science that went into what appears to be a simple product, like the toilet paper, toothbrushes, or soda cans like you mentioned.

[Walter Roberson]

As an example of high tech science in consumer products: Pringles used Cray computers to model the aerodynamics of the chips traveling through the manufacturing line, because the chips travel so quickly through the process that for parts of the process they are traveling on a cushion of air.

[Sven]

Walter, did you happen to go to the exact same LS-Dyna conference I went to... or does P&G give a similar presentation in a number of places :)

Diapers, stackable drink/detergent containers, food production lines... all topics heavily dependent on both image processing and mechanical/fluid dynamics. I have to admit, the pringles analysis resonated with me too.

IA, do we have a chance of guessing "the" product you were talking about, or were you talking about a suite of products made by someone like P&G. And if it's top secret military stuff, don't worry, you can trust the internet to keep a secret ;)

[Walter Roberson]

SGI bought Cray, so the Pringles item appeared in the SGI publicity at a time that I did a lot of work on SGI systems.

Sven, IA is the person responsible for determining what color it should be.

[Image Analyst]

I work on an extremely wide variety of products. I haven't worked for the military since 1981 (as a civilian). I might or might not work for P&G. The information I want to reveal is in my profile. By the way, P&G sold Pringles to Kellogg's about a year ago.

[kazi]

for everyone i am working on driver fatigue detection......that's why i need to detect yawn that's mean mouth condition or state....please anyone help me about it

[Image Analyst]

Like I said, I hope you're using thermal images, but I doubt it since a thermal camera would be too expensive to add to a car. But anyway, where did you upload your set of normal and yawning images?

[Walter Roberson]

How about one of those CMOS based sensors with the infrared blocking filter replaced by one that selects mostly infrared ? Like the infamous Sony handicam that could do infrared?

[Sven]

Oh, and Walter, I've been collecting deciduous forests for years now in the hopes of one day purchasing a ship's peanut.

[Image Analyst]

I don't know if a CCD sensor can detect an open mouth. The sensitivity of CCDs ends at about 1000 nm (1 micron). That's near IR. The thermal IR range is way way out there - up around 8-12 microns - far past what a CCD can see. If you're looking for temperature differences then I don't think a silicon CCD can do it.

Answer 1

[Walter Roberson]

http://www.mathworks.co.uk/matlabcentral/answers/37600-eye-recognition

[kazi]

why you exactly need eye recognition? as you mentioned you can detect eye from face.

[Walter Roberson]

"I was making a project to detect driver drowsiness [...]" and Image Analyst replied indicating a resource to examine for algorithms.

[Walter Roberson]

You never did post links to "normal" and "yawning" images.