Thread: REM Detectors (again)

i've been reading up on the 555 timer lately and it might do the trick (i think).

there are two things we could measure, and i don't know which would be the best way to go:

1- the difference in skin resistance to a "fixed" point ("fixed" meaning it can be calibrated) -> 555 timer -> serial port (i don't know much about the serial port, so i'll have to look this up a bit more)
2- the true value of skin resistance (which would directly control the 555 timer... in which case the calibration would be software based)

i think the software based calibration is the better idea (given that we're aiming for as few components as possible to begin with), but i don't know how reliable the hardware will be (given that, depending on the construction and placement of the probes the resistance will be different)

i'll do a bit more research on the 555 timer (i really thought the voltage -> frequency was a good idea, but this seems so much easier...) i know i'm going through ideas and approaches faster than i can actually build any of them, but i think this is promising.

i'll start by building a simple led flasher and i'll play with the capacitor values. the idea is to use my own skin resistance to flash the led (with a fixed 250 kOhm resistor, a 0.1 uF capacitor and a variable skin resistance between 10 kOhm and 500 kOhm, we get a range in frequency from 12 Hz to 65 Hz... a soundcard would not pick that up, most skrew up under 20 hz or so, but for serial port it should be ok... if we need more accuracy (which we probably will), we can boost that up in the kHz so that we get a bigger range...

for testing purposes, i'll try to keep it even lower than 12 Hz so that i can see the LED blink.



if this works, we'll have a very very simple device to build. i don't know if the 555 connects directly to the serial or if it needs some sort of converter (does anyone have a quick answer/schematic that may save up on searching?)

kimpossible, since you used 6 subjects so far, does the resistance level vary a lot across individuals, or is it somewhat similar? also, given your experience in electronics, if you have any comments on what i just said (am i completely off anywhere ?), it would be greately appreciated (actually any kind of help would be appreciated)

oh yeah, the only small problem that may add some components is the voltage value we're willing to run across the skin... from what i read 0.5 V was the optimal value, so i'd have to figure out how to make that work with the 5 V of the 555, but it shouldn't be a big deal.

Hmm, looks like we're gonna end up with two prototypes
Welcome back Meus

i have all the part for the 555 timer circuit, i'll be experimenting tonight with with a LED. I'm working both tomorow and the day after, but I should be able to tell you how it goes pretty soon.

The 555 is pretty antiquated. I think you'll find that you can't keep it from drifting to the point of being useless.

Time to start thinking about looking at a microprocessor - like a PIC.

actually the 555 seems to do the trick for now. I wired up a basic led flashing circuit, and with a 200 K resistor and my skin resistance (about 200-250 K) i see variations in the flashing rate as I hold the "probes". I'm using a 1 uF capacitor and a large resistor, in order to see the led flash (with a 1K resistor i can't see the flashes anymore, and with a 5K i can make out some very quick flickering)... if we use a 1k resistor and 0.1 uF we should have a high enough frequency to provide for (somewhat?) accurate readings...

i WOULD use a PIC and I realize how much easier and more accurate it would be, but... i don't know that much about electronics or programing, and i don't have the required time to learn these things.

the schematic i used was the astable one on this page: http://ourworld.compuserve.com/homepages/B..._Bowden/555.htm
I used R1 - skin, R2 - 200 K and a 1uF Capacitor, which gives (with my skin) a reading of about 2Hz. My girlfriend however has twice my skin resistance (in the 450+ K range) which takes the frequency down to about 1 Hz
to wire up the led, i just connected a 470 ohm resistor from pin 3 to a led. (pin 3 to resistor to + LED).

the problem right now is that we use about 3-4V across the skin... it's not a lot, but i don't know about long term effects of that. I'm using a 9V battery, and there are a few ways to go around it, with the easiest being a voltage regulator... if i drop the voltage to 5V, we'd get some 2-3 V across the skin, which is slightly better and which the timer can take (some timers apparently can go down to 3V, but i don't have one of those, and i don't know about their availability... i want to keep this as simple as possible.)

anyways, if anyone's willing to help connect this to a computer, it would be great... if not, i should be able to figure it out eventually.

next steps:

- small capacitor (1 or 0.1) across skin to filter out noise (i have to buy some 0.1 uF ones cause the lowest i have is 1)
- 5v voltage regulator (i have one)
- figure out how to connect it to a computer and get some readings. (given that the output is either ON or OFF, i figure this should be easy, but i don't know anything about the serial port... i'd go with the line in option on the soundcard, but now i think this would just complicate things for nothing)

so yeah

"see the difference in flashing"

You realize we're talking about fractions of 1% for the range of difference between rem and sleep, right?

no i don't actually... that's why we're trying to build it, so we can record a few nights sleep and see if it offers a simple alternative to eye movement. from what i read, the shift in skin resistance was easily noticable, but my sources may be wrong, or i may have misunderstood the articles i've read.

but instead of opting for elusive replies, given your knowledge in the field, i don't see why you don't try to help us out. there's no reason why you can't keep your existing design, comercially produce it, but also help with ours...

Sorry - I can help point you in the right direction, but I'm not going to gut my own devices by designing another for you. That'd be kinda silly.

I thought I was giving you some pretty clear pointers, actually...

In order to see enough delta for reasonable self-calibration, I needed to amplify the signal more than thirty-fold. At that kinda SNR, you're going to need something to take the edges off.

Also, what are you looking at for a sample-rate?


As far as asking for my comments: I offered some. The 555 drifts significantly. You pushed that aside "oh, it's working fine for me" but you don't have any measurable data to confirm that. The actual drift is about 1% with another +-1% for the supply. If you figure another few percent for the voltage regulator, you're going to have calibration issues.


Skin resistance varies from person to person. In my tests, I've seen my own vary throughout the night from 1.3MOhm to 105kOhm. The range I've seen on my test subjects is from 52kOhm to 3.3MOhm.

well, the idea was that all the necessary calculations could be software based for now... the immediate aim is to get a raw signal (even if it's really noisy) into the computer and see what it gives us... this way we'd know how much we need to clean it up, how much should be hardware-based and how much can be software.

there is no real need for amplification at hardware level (at least not yet, i don't know what will happen once we get some readings), because knowing the capacitor and the resistor values, we could estimate a somewhat acurate value for the skin resistance (especially if the 555 frequency can be recorded at two or three decimal numbers) This would be used for the calibration (which again would be software based)... of course, the accuracy of the 555, resistor and capacitor will play a big role, but i hope the difference will be fairly noticable, else we're just looking at problems, starting with the home-made probes (which can't be that constant or accurate to begin with)...

the sample rate could also be software controlled... if i get a pulse to the serial port, varying from say 20 hz to 60 hz we can figure out at what rate we need to check the information... i don't know how fast a serial port is, but i trust a pentium 4 can perform more than just the necessary calculations in real time and decide from there what values need to be taken into consideration for REM detection (average over a few seconds, or whatever)... the software could also probably provide the necessary amplification... if the overall shift in skin resistance is say only 30 K, which would give us... let's say, a 10 Hz range (at two or three decimal numbers, or however much the computer will allow)), we'll have at least 1000 points to play with on a time graph, which should be, again, fairly accurate... if the shift is less than that, we're probably running into trouble...

the idea is that rather than amplifying the signal and trying to get rid of noise and playing with very small values for GSR detection, i'd rather just scrap it and pick another feature that may provide fairly accurate REM detection, like breath rhythm/intensity measured with a strap across the chest... and if that's not enough either, maybe even combine two less reliable detection methods.

what i'm interested in is the easiest, cheapest way to (somewhat reliably) detect REM... even if i get a 50% detection rate, i'd be satisfied... it means 50% more chances of having something signal to you... which is way, WAY better than nothing . Either ways, even if the REM detection is not reliable enough, the unit could still be used as a biofeedback/meditation tool.

so yeah, for now the idea is to get SOMETHING into the computer, however unreliable it may be, which is why i need to figure out how the serial port works

Originally posted by kimpossible
As far as asking for my comments: I offered some. The 555 drifts significantly. You pushed that aside \"oh, it's working fine for me\" but you don't have any measurable data to confirm that.

I didn't push it aside, i said it works for now (i got a LED to flash, which I didn't have the night before hehe ) ... it might not be very reliable, but i wouldn't know about it it. That's why i need to build it, to have measurable data (which, as you pointed out, i don't have yet). I don't know if it's going to work or not... but i do know i have no idea how to make a PIC work... which sums up to: if simple electronics won't work for me, i can't build it.

As for skin resistance... a 3 MOhm reading would probably be very poor in terms of accuracy (the smaller the resistance, the better the reading with the timer), but maybe that can be fixed with some variable resistors for calibration further on down the road.

Either ways, i'm doing this for fun... if it works, i'll be delighted and i'm sure some people will find it useful... but if it doesn't, i have nothing to lose (other than a few $ and a bit of time).

and i'm not trying to be mean by saying you could contribute a bit more, but i honestly think it wouldn't take away from your target market... there are plans for the Kvasar out, and i don't know many people who built one... in fact, there are people who'd pay close to the price of an REM Dreamer or a Nova Dreamer to have it built for them.
I certainly do appreciate the input you're offering (especially in your last post, considering skin resistance values).

Actually, she did contribute quite a lot, until recently when she removed her posts.
Her prototype thus far is very successful. I just hadn't heard much about it in a long time.

i probably missed all of the posts that are now gone... aaanyway.

i was looking at the plans for the dream weaver (i'm planning to build one too, i just ordered the specific led/phototransistor part that i needed, along with replacement parts for my kvasar, to see if i can pick that up again) and i realized the game port might be, for the time being, an easy alternative to the serial (till i figure that one out... it seems to be fairly complicated compared to the game or the parallel port).... i'll be visiting my parents soon and i'll pick up my old laptop (the one i have right now doesn't have a game or lpt port), which is a 486 that i'm not afraid to experiment on by plugging in all sorts of random home-grown products it's limited to dos/win 3.1, but from what i hear win xp comes with all sorts of restrictions on the ports that we'd have to work around (so i have to figure that out too)...

the plans for the dream weaver are here: http://www.cerebrex.com/0ldGraphics/drwevr.zip

from an ld4all post, the sensor can be replaced with a CNY70 and the 2N22222 is drawm wrong, with the emitter and collector mixed up... but the game port reads straight from the phototransistor, meaning i could probably hook up a transistor to the 555 timer and plug it in ? i'll have to look at it a bit closer, but it might do the trick.

Oh my god. I havent noticed this thread. Actually I havent given it any importance since it is in the research section. I thought the reasearch is only for lucid gods. But this is really interesting and it is my subject.

I was designing a circuit to detect REM using a different method. Wait.. I have to read all the posts !!!! ( I wont be patient enough to read all those previous posts. If someone can post a summary, it will be nice.)

What about this experiment? Have you find out something? I'm thinking another idea that also includes this one. I thought about a NovaDreamer but sounds, lights, touchs... and all in a REM state.

I would like to know (if possible) if you have discovered something. My idea about the REM detector is this: A timer will wait until you are sleep and near the first REM. That is, timer = time you need to fall asleep + time since falling asleep to first REM.

I got some electronic knowledgment, if you need help with timers (those 555) or connecting it to the computer just let me know.

This project is on hold indefinitely, at least for me.
Perhaps one day I'll pick it up and continue. Perhaps not.
I get disillusioned when the people with all the data and prototyping decide to F off with it all...

My electronics is limited but I do remember 2 basic facts you could use to your advantage.

The inductance of a coil changes when a body part gets close to it.
The capacitance of two plates also changes if your interact with it.

The eye is not symmetrical so, in theory, it could affect the fields when moving.

Using the inductor technique:
place a small coil directly in front of one eye. Close to it. Measure the inductance. When it fluctuates for more than a minute, you got rem.

Using the capacitance technique:
place a small plate on both sides of one eye. Measure the capacitance of the crude capacitor. When it fluctuates for more than a minute, you got rem.

Pretty good ideas icuurd12b42.

I was thinking in another way. Some people (me for example) couldn't sleep with something similar to a mask in their faces, it's horrible for some of us! Another thing is that some people feeling bad when they think about some IR over the face for so long (I can say that that IR leds are non harmful at all)

My idea was to use those IR with a temporizer. The temporizer will activate the device when the REM state is suppossed to appear (a little before 90 minutes, am I wrong?) and it will activate a "robotic-arm" (let's call it a stick with the IR on it) that will go next to the face of the dreamer and tries to detect the REM state for a period of time. I've opened another topic with an application for this device where it's supossed to interact with the dreamer giving him some stimuli (not only visual)

The problem with detecting REM using the eye is that you have to follow the eye. A mask does this naturally. Until it falls off or you tear it of while sleeping that is.

An unattached device would have to follow the face of the sleeper as he switches side and position.

If your device is unattached though, you may be more resourceful as many options are now available. Remote Face tracking and REM detection is possible using a CCD camera on a robotic arm and computer though the concept seem very costly.

I think the cheap solution is a NovaDreamer type mask with coil inductor to detect REM.

I used the NovaDreamer myself. I found the LEDs used to interact with you were way too bright. I would use a vibrating pin on the forehead instead. If you keep getting bit on the forehead by mosquitoes, you know you're dreaming.

The electronics inside the NovaDreamer are so old school, you could have made one 30 years ago.

In the same size package, you could have a micro computer. A pic as mentioned. To drive the detection coil's data acquisition for REM detection. Once detected, drive the vibrating pin or LCD to make you aware. The awareness method does not need to be fancy. I believe the more complicate it is, the more convoluted it would be in the dream, making you less aware as the tool would turn on you and start driving the dream as opposed to making you aware.

I also liked the idea of the glove mentioned here. Sort of reminded me of a ST Voyager episode.

Originally posted by icuurd12b42
The problem with detecting REM using the eye is that you have to follow the eye. A mask does this naturally. Until it falls off or you tear it of while sleeping that is.

An unattached device would have to follow the face of the sleeper as he switches side and position.

If your device is unattached though, you may be more resourceful as many options are now available. Remote Face tracking and REM detection is possible using a CCD camera on a robotic arm and computer though the concept seem very costly.

I think the cheap solution is a NovaDreamer type mask with coil inductor to detect REM.

I used the NovaDreamer myself. I found the LEDs used to interact with you were way too bright. I would use a vibrating pin on the forehead instead. If you keep getting bit on the forehead by mosquitoes, you know you're dreaming. *

The electronics inside the NovaDreamer are so old school, you could have made one 30 years ago.

In the same size package, you could have a micro computer. A pic as mentioned. To drive the detection coil's data acquisition for REM detection. Once detected, drive the vibrating pin or LCD to make you aware. The awareness method does not need to be fancy. I believe the more complicate it is, the more convoluted it would be in the dream, making you less aware as the tool would turn on you and start driving the dream as opposed to making you aware.

I also liked the idea of the glove mentioned here. Sort of reminded me of a ST Voyager episode.

Agreed, I'm surprised nobody has updated that circuit board. Its huge for today's technology. I can't believe they were charging what they were charging ($699) for such a huge piece of old technology. If someone really invested in new development I'm sure the new one's would be much lighter and comfortable.

Originally posted by Placebo
I get disillusioned when the people with all the data and prototyping decide to F off with it all...

I hear you there Placebo! That person was going to market the product and make gazillions of dollars on it. So far nothing, makes me wonder if they really did it or were just making it all up. I bet we have a backup of it around here somewhere. If so, I'll see about restoring it.

Originally posted by phoenelai


Agreed, I'm surprised nobody has updated that circuit board. Its huge for today's technology. I can't believe they were charging what they were charging ($699) for such a huge piece of old technology. If someone really invested in new development I'm sure the new one's would be much lighter and comfortable.

Wow. did you pay that much? I was lucky to get it on sale. When I looked at it I was a bit discussed with the 5$ circuit board. On the other hand, cuicuitry aside, It would have cost me more that what I paid to build one. Integration to mask and final prototyping is costly when you only build one.

I got it in a package. When I counted all I got with the mask, I was not too angry with the purchase. The package was 250$ if I remember right. I figured the mask, in the package, I paid the equivalent of 150 bucks for it.

I think both the coil and capacitance ideas have the downside of needing some sort of mask, where as the brilliance of the GSR circuit was that it would be non-intrusive. So is there any reason to think more about them - like simple design?

Here's a project where they build a very small GSR. Unfortunately, not all the data is there so it's impossible to use it as it is.

http://bentham.k2.t.u-tokyo.ac.jp/notebook/?cat=3

Here is some other GSR construction link:

http://web.media.mit.edu/~carsonr/pdf/gsr.pdf

Thank you, I thought the unintrusive glove was an excellent idea too.
Maybe I'll try to make it again one day. Maybe.
I'm no electronics whizz