Dear Dariusz.

You can find at least two products with several options and prices (http://www.smarteye.se/ and http://www.seeingmachines.com/). I am not sure about http://www.noldus.com but you can try it.

You talk about driver fatigue detection but there are other works related with drowsiness and distractions. They are different but related problems.

In my humble opinion, there are several important facts in this problem.

Experts in traffic safety are needed to provide ground truth information.

Based on our research in automatic detection of driver distractions, to obtain a robust risky signal, you will need to take into account information about road, vehicle and driver (we combine all of them using a fusion method). We used a camera to detect driver´s hands activity. In some experiments we also considered driver eye information. At the end, driver behavior is a very complex problem and a multi-disciplinary approach is required (in our case: Traffic Safety and Technology experts).

Hope this helps,

Enrique

I'm inclined to agree. Eye tracking solutions are a dime a dozen. You can build a very simple one in OpenCV or MATLAB with using some provided code.... things like checking the frequency of how your eyes are blinking, or how your eyes are dilated can be factored into the system as well.... but personally, I would not pay that much for an eye tracker. What's probably factoring into the price are all the frills, bells and whistles, like using up to date technology and making the system as small as possible.

They're most likely banking on the fact that a lot of people aren't familiar with the Computer Vision area, and they're beating other people to the punch first to actually make a solution that wow's the public.

Also, there's a very simple solution to combating eye fatigue.... don't drive when you're tired!

I feel $20K is indeed too expensive. The production cost should not be such expensive, the viable system should not cost more than $500.

Dear Dariusz.

You can find at least two products with several options and prices (http://www.smarteye.se/ and http://www.seeingmachines.com/). I am not sure about http://www.noldus.com but you can try it.

You talk about driver fatigue detection but there are other works related with drowsiness and distractions. They are different but related problems.

In my humble opinion, there are several important facts in this problem.

Experts in traffic safety are needed to provide ground truth information.

Based on our research in automatic detection of driver distractions, to obtain a robust risky signal, you will need to take into account information about road, vehicle and driver (we combine all of them using a fusion method). We used a camera to detect driver´s hands activity. In some experiments we also considered driver eye information. At the end, driver behavior is a very complex problem and a multi-disciplinary approach is required (in our case: Traffic Safety and Technology experts).

Hope this helps,

Enrique

Thanks for the feedback everyone!

@Enriquie: I appreciate the links and info about the related problems. I was wondering if you could be more specific about the road info for your robust signal. Are you referring to the 'actual' road or are you being more general and referring to environment (i.e. other vehicles, pedestrians, obstacles, weather, etc)

@Raymond: Yes, I think the bells and whistles, like tracking a driver's history and reporting it to management, definitely up the price - and perhaps for a mining company, $20K per vehicle is a bargin to protect against the liability of tired drivers.

Prices are often decided in a way to maximize profit, selling few devices at high prices may be optimal strategy.

If your goal, however, is not profit, but to save as many lives as possible than a non-profit organisation with real mass production and low prices, maybe even with subsidised by the government may be the optimal pricing strategy.

In my opinion, such kind of system should work under IR illumination. In other case, you only can work during day hours and light and shadows will degrade your system performance. You should also consider if you need a system for a car or for a simulator. I think commercial systems provide these elements. If you would like to do it, it is possible and not very difficult for an expert Computer Vision programmer. If you are involved in a project or you are a company, you should evaluate programming effort, testing time, hardware costs and your environment to know if you should do it or buy a product.

Regarding road information. As the statistics expert in my division say: as much information as possible.

If you use a simulator, all information is available (distance with other cars, etc). If you are in a real car, you can obtain information about road signals (speed limits, curves), you can measure the slope of the road, you can detect you position in the traffic lane (detecting lane limits), you can detect if there is a car or an obstacle in front of you.

This document reviews the issues and describes more than 20 devices:

http://www.fmig.org/CATFatigue%20Technology%20report%202008.pdf

Eye tracking ≠ fatigue detection, which works from eye movements (and subtler face indications), by recognising more complex events, and recognising the ones highly correlated with fatigue. Even at the lowest level, do the 'dime a dozen' eye trackers report blink rate?

I am sure the price will come down as others get into the market, because the hardware costs a lot less than $20K: but right now, validated fatigue detectors do not grow on trees. A company considering the costs and liability from a fatigue-induced crash could find this price worth while.

in my opinion, no it is no expensive. Inversely, it is cheap

It's variable... if this is a critical function, which also demands precision, it's not expensive. Else, i developed myself something like that with Actionscript. There's examples over web in many languages, in this case, $20000 is expensive.

About Caterpillar's case, it's not expensive because this is a critical function that takes responsability. Is must work. Always.

To develop a new brand for a little store costs only $3000, but if this is as big as wallmart, will cost much more. This is real in Design area, and this concept must be considered in softwares.

"Newmont Mining told the BBC it believed DSS had reduced fatigue-related incidents by 90% during a pilot study at one of its sites in Nevada." According to this quote, it still misses some fatigued drivers, and it is unclear what the typical frequency for fatigue-related incidents is.

I am curious if perhaps the drivers "felt" more awake because they believed the system was watching them. I wonder what would be the results if someone installed a placebo system that looked like the eye tracker, but in reality wasn't connected to anything.

If the face-tracking system issues warnings and alerts, a placebo would be difficult.

Dear Dariusz good point - though I wonder if this "experimenter effect" would not already be a great solution to the problem. Sometimes it can be so easy.. On a more serious note, I agree that this is information you need, to judge how well such a system really works. And indeed, I believe that the technology to implement such systems exists - also in the form of physiological sensors that are able to monitor mental states independent of overt behaviour. Of course, as already mentioned before - to solve all the implied challenges (from hardware - robust sensors that can deal with the extreme environment - to software level - classification approaches that reliably run under the same extreme circumstances) is not straightforward and needs many clever (expensive) people.

Dear Tim - could a placebo system not simply produce warnings in a random manner. I do not see a problem to compare a random system with an actual working system to make drivers in both cases more careful as they feel watched - and then, on top of that, to compare the effect of adequate warnings versus simply annoying disturbances.

I know something of my body and fatigue, and symptoms like yawning, or a blink turning into a longer closing of the eyes. A system that caught those would impress me (whether or not it made me a better driver), while random warnings that missed them would be obvious rubbish, making a placebo useless.

Of course, no system will ever make me a better driver unless and until I learn to drive a car, which I have no intention of doing.

I think to solve the realistic problem of fatigue detection is hard. If the system is precise, robust and works under all conditions, no matter day and night, it should worth the money, otherwise, it's too expensive.

I believe there might be other kinds of techniques to detect drowsiness.

actually, I found that it is possible to detect drowsiness based on 1chanell EEG signals and ANNs with results: 86.5 % for drowsiness stages and 81.7% for alertness segment .

I think dry electrode EEG devices might be considered as possible solution.

U can find some patents and articles about this topic.

developing EEG systems may decrease the costs.

Dry electrodes in a sweaty cab? Problems even before you start thinking about driver consent to wear them -- and carefully fit them on whenever they get set to drive.

Some countries still have trade unions.

Well, difficult to say whether it's worth the price. it mainly depends on what level of accuracy you are aiming to. If you want something that *really* works, it's probably a reasonable price. I am the co-founder of a small startup company working on eye-tracking and you can use the information captured by the sensorial equipment at very different levels. Blinking is different from tracking, and also from fatigue detection.

The nuia eyecharm project is very interesting, although so far the results presented are rather limited. It would be interesting to see what the final result will be.

if I'm not wrong also car manufacturers (Mercedes) started releasing some Attention Assist packages as optionals on their cars. no idea what the price can be. If you think it could save yours and others' life, it may be worth the price...

Cheers!

Alexey, take it easy :-)

First of all I said it MAY, which means that it depends on several factors. Apart from that, the system you mention by Ford, the Driver Alert System, does not rely on any eyetracking technology. it monitors fatigue by checking if your driving patterns are stable or not, mainly in terms of lane detection, which is not viable when you don't have lanes, as I suppose it happens, for example, to Caterpillar.

Cheers!

Not only eye fatigue,but Leg Ache,Hand pain,Cramps,agitated mind,Drunken state,Diversion of attention due to Chatting while driving,use of mobile phone,viewing of T.V or Videos on cell phones etc.while driving can all lead to accidents.Self analysis of fitness ,and sticking to Safety Rules is the best way for safe driving or any other job that needs concentration.

If we go on devising sensors and controls for every such condition there will be no end to cost escalation.

Such devices are useful to find whether it is technically feasible and will be useful when they are economically viable.

An Engineer's job is to design a Useful Device that is economically viable.

Robots,I suppose, do not get fatigued as easily as a human beings and do not require such a device.

This like using a Microprocessor or Microcontroller for every job,even if it can be done easily with a cute little Hardware Device.

P.S.

My guess would be that any attempt to sense or detect the state of the driver is going to pose endless problems. It makes far more sense to equip the automobile with sensors that would detect erratic movements of the vehicle that would indicate that the driver isn't in full control of it. Perhaps your point is to detect the possibility of this before it happens but we humans are just so unpredictable. I can't imagine any way to truly detect the state of the driver that wouldn't involve connecting all sorts of nodes and monitors which becomes very unreliable due to those who just won't do it.

Alexsey, "if it is about wear or lose the job - the driver's opinion becomes not so important" is true if the driver stands alone, but as I pointed out, "Some countries still have trade unions."

When the union's inspiration through the workers' blood shall run

There can be no power greater anywhere beneath the sun

What force on earth is weaker than the feeble strength of one?

But the Union makes us strong.

Micromanagement by machine just might make that message resonate again... until the machines take over the driving too.

Trade unions have the power of the strike, which is not always exercised in the interests of rational scientific management. It was rational management that used to fire workers in the Yorkshire coal fields when their hair turned grey, because they were less productive: and this was before pensions. They used coal dust to darken their hair. The unions made their lives better, forcing safer conditions and better pay.

If the machines track what you are doing, every second of every minute, they will not only save lives. They will be used to limit your time on the toliet, or talking with your friends, or just collecting your wits before the next bit of work, all in the name of shareholder value. I promise you, some people will resist this.

Even some people wearing Google Glass will resist having it report every moment to the boss machine.

Even when there are laws against strikes, the union can refuse if its members are willing to fill the jails -- as they sometimes have been. The unions are often corrupt old organisations now, without that kind of solidarity, but enough oppression will bring it back. As somebody remarked three thousand years ago, "The thing that has been, it is that which shall be; and that which is done is that which shall be done: and there is no new thing under the sun."

Dariusz, there has been a lot of research in the field of driver intent from Trivedi, et al at UCSD. Seems that a particular driver's intent can be fairly well learned over time with some reasonably priced sensor suites. From having the knowledge of a drivers intent in certain scenarios, it may be inferred, that if a behavior occurs out of the known reactions for that driver/scenario, fatigue may be one of the causes for such behavior. Here's a list of their papers, many of which I have found useful:

http://cvrr.ucsd.edu/LISA/Publications.htm

Caterpillar is also looking into vehicle autonomy for some of their fleet, exposing other driving problems..:)

I would say $20K is too expensive for most commercial applications. We are beginning to see smartphone-based applications that are reliably performing the same capability at a fraction of the cost.

Yes, a trade union may oppose anything that its members resent: that is a part of democracy. Electrode scanning is more invasive, disturbs the hair do (for those with more hair than me), tickles, has to be close to the skin for accuracy, very uncomfortable in sweaty conditions: many people will hate it. It also requires active co-operation by the driver, in putting it on, which allows passive resistance short of a strike -- the drivers can agree that none of them will put it on. Can the company afford to fire them all? Fighting an eye tracker 'spy in the cab' needs an actual strike or sabotage of the installed equipment, which are bigger steps, though not impossible. (Compare http://archive.commercialmotor.com/article/18th-october-1968/31/spy-in-the-cab-strike.)

Many motor cyclists resent helmet laws, and many car drivers resisted seat belts. (In Israel you could buy T-shirts with printed diagonal straps, so to a police glance they _seemed_ to be compliant.) Those people are not going to like scalp electrodes.

There may be, perhaps, on balance, a rational case for them: for belts and helmets, there certainly is, despite libertarian objections to their enforcement. But there was a fight to make those normal, and this fight would be certainly be a bigger one.

I agree that seat belts are wise, and I do not oppose them: but I was pointing out that many people did oppose them. That is a fact, regardless of whether you or I think they were acting nonsensically.

This fact answers your question "would actually anyone object wearing it?" People objected to the point of holding a strike against the 'spy in the cab' , and that was physically less intrusive than electrodes or even Glass. It was merely something installed in the vehicle, like the facial fatigue sensor this thread was originally about. Comfort was not the issue (though discomfort could add to it)

It was extremely intrusive on people's actions, and this is what people objected to. Whether they were right to oppose it is another question. Whether it is justified to object to fatigue detectors (and all the other information the company would collect with them) is one question. Whether people _will_ object is a quite different question. Some people will, and the more physically intrusive the detectors are, and the more that people have to co-operate by putting equipment on themselves, the more objections there will be, and the more opportunities to prevent it from working. Companies that casually assume "nobody will object or sabotage this" will create big problems for themselves.

I did not say "no chance", only that it is likely to be opposed by the workers. Worker opposition sometimes succeeds, sometimes fails, but a company that assumes there will be no opposition and puts no effort into worker buy-in will be unprepared for a nasty fight: even if it wins, it will emerge with damaged morale and worker loyalty.

If I were a driver I would accept fatigue sensing (via eye tracking or by scalp) only if it is narrowly restricted to its life-saving functions, like warning of slowed reflexes and requiring a rest (with full pay). If it is used (for instance) to dock the pay of a father who comes to work tired after a night with a sick baby, you can expect mass anger and revolt.

According to my study experience in mainland China, manufacturers or solution providers can offer a 2000 USD price per set proposal.

Drowsiness is harder to control consciously than other factors - that makes detection important.

Detecting erratic driving behavior - may be too late.

My vote for eye detection. The price will go down quickly, as with any new technology.

Interesting idea/product. Eye monitoring using cameras seem to be better than multilead EEG systems (cap electrodes etc). However, I have a few questions:

Is there a finite latency in decision making for this system?

Are there multiple levels of fatigue/alertness measures? What is the probability of false alarms?

In turn, is the device having multiple severity in prompting? what if the driver does not react to prompting?

Are there other more technical presentations about this device?

I tell you that the eye tracker can be order 80,000 USD, however it is important to look at the potential of the added software and its applications in different areas of knowledge

Initial systems that are trying to bridge the chasm can come out at an overly expensive price point like this system. The thing that people will have to look at is what value does this system bring? Will it save lives? will it save money? will it make for safer roads? The algorithms and hardware probably already have millions of dollars of R&D money that the company is trying trying to recoup. If this product catches on the price will fall as other industries require the tracking module. Right now for the common man and truck driver it is by far too expensive, but for large companies and mines it could save millions of dollars making it a valuable resource.

Given the features,

1.uses a camera to detect a driver's pupil size

2.truck cabs are also fitted with an infrared lamp

3. An accelerometer and GPS chip are also installed

4. it triggers an audio alarm and vibrates a motor

5. An alert is also sent back to the miner's support staff

6. video feed of the driver's eyes and view data about their recent behavior.

the price is rather fair. It's human life we are talking about and system is currently focusing on truck drivers who driver for consequent hours without rest. If it succeeds, then like all other systems, hardware can integrate and price will drop down slowly.

And it's like it will be compulsory like seat-belt until the price is rational and system accuracy is dependable. type of alarm can be adjusted, other factors are more important.

Without adding to much into what sensor modality is best to detect drowsiness of drivers (which appears to be a direction this conversation has spun into), if your just after a cheap eye-tracker sensor and are willing to do research on top of it then https://theeyetribe.com/ maybe worthwhile.

Have not had the sensor personally, came recommended by a colleague that was looking into detection of drowsiness (using EEG).

Ridiculously expensive. Driver fatigue can be detected via sensing of steering input behavior (which is "free" since modern cars have steering angle sensing as standard).

It is already done, validated and in production - the characteristics of the driver's steering correction input change (eg longer delays and larger corrections). No so different from estimating from blink rate (eg less frequent but longer blinks).

Peter, you might want to ensure that you are fully alert before commenting....

Yes, $20k for that is too expensive.

Although it seems expensive now, in a few years this technology will create a market demand and become affordable, just like PCs and the like. Safety for all!

We have implemented a simple, but efficient algorithm based on FFT value of the steering wheel angle whose precision is far beyond 70%. See Tobias Islinger, Thorsten Köhler, and Bernd Ludwig,"Driver Distraction Analysis based on FFT of steering wheel angle", Automotive User Interfaces 2011, Adjunct Proceedings, Salzburg, Austria http://www.auto-ui.org/11/docs/AUI2011_adjunctproceedings.pdf if you are interested.

Coming back to the initial question, we are using a mobile eye tracker by SMI (http://www.smivision.com/en.html) which costs around the same price as the catapillar system. So, in my view, for an eye tracker the price is fair.

Our work was not on fatigue detection, but distraction detection. Our results indicate that when drivers perform a secondary task in parallel to the primary task of driving the car, the steering angle is a good predictor for a secondary task.

There exist lots of eye tracking solutions (both commercial and non-commercial) that can be used for not only fatigue detection but also eye-controlled tasks. You can take a look at them. Besides, most of them have affordable costs that are far lower than 20.000 USD.

You may start by reading this article:

http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=6517478

or

http://www.researchgate.net/publication/237148063_A_Low-Computational_Approach_on_Gaze_Estimation_With_Eye_Touch_System

Article A Low-Computational Approach on Gaze Estimation With Eye Touch System

If it is for the purpose of contributing to advance scientific research in this area than it is not expensive because in the future this price will be very low with more advancement in the technology.

$150 tracker from http://www.smootheye.com may be used for the purpose