Sensor to detect Human Presence

Question

Hello Friends , Is there any sensor to detect human presence . Am aware of PIR ( Passive Infra Red ) sensor , which would human motion but not the human presence . Another way is by using image processing technique. Kindly suggest me a better solution . Thanks.

Former Member · Answer

Hi Gunaseelan, OMRON do a thermal sensor that can detect stationary thermal objects as opposed to standard PIR motion sensor that only detect motion. The OMRON D6T thermal sensor is able to detect the presence of stationary humans by detecting body heat, and can therefore be used to automatically switch off unnecessary lighting, air conditioning, etc. when people are not present (regardless of whether they move or not).

beacon_dave · Answer

shabaz Voight-Kampff machine ? Voight-Kampff machine | Off-world: The Blade Runner Wiki | FANDOM powered by Wikia

BigG · Answer

This sensing technology just crossed my radar (or "field of view"). https://www.xethru.com/ https://www.youtube.com/watch?v=qzJvhfIwZQo Looks pretty good. No idea of price / performance though.

shabaz · Answer

Agree.. I don't know why the OP excludes image processing - due to incorrectly perceived difficulty, or for a non-technical reason.

I can understand the desire to possibly exclude due to (say) security/sensitive locations or countries with laws on what is allowed to be captured/temporarily stored, but there are ways to work around that - in particular, replace the video camera with a far lower-res thermal camera imager. For a project I worked on, we did exactly that, using a fancy FLIR Lepton sensor - eliminate the awkward conversation with customers about capturing video in on fell swoop.

shabaz · Answer

Hi Mrigank, Perhaps I'm wrong, but anything with a scientific quality and quantity, such as physical, electrical, chemical attributes, can be measured - might not be easy, might be constraints, and might take expensive equipment, but it can be done. However, you need to examine your real-life requirement (e.g. "emotions") or "human presence" as the original poster wanted, and translate it into the particular scientific attributes you wish to observe or measure. The ones you choose can be constrained to your requirement. For example, if you know that you are only interested in the emotions of people prone to outbursts when they are standing at a specific quiet location, then that is a constraint you can make use of, and then you could use a microphone placed at that location as a sensor because the scientific attribute you wish to measure could be 'sound level'. Or if the emotion results in turning red and hand-waving then perhaps the best sensor is a camera : ) I've no idea. If you can do some research, examine if you can successfully obtain your required result (a gauge of emotions as you say) using (say) a measure of rate of heartbeat, and if so, then a technical discussion can begin on how to measure the attribute such as the rate of heart beat. Whether heart rate corresponds to emotions I can only speculate - perhaps a psychologist would know the answer - no idea I'm afraid : (

dougw · Answer

Sometimes it requires a close examination by a doctor to determine if someone is alive. But you could try this: https://www.omron.com/media/press/2013/09/e0917.html https://www.element14.com/community/message/229828/l/looking-a-for-facial-recognition-hardware-tester#229828

nschreiber0813 · Answer

Dear: Gunaseelan If you are trying to build a human presence sensor there are a few things you should know about that. First of all humans tend to have certain things that are detectable like for instance... Motion Heat CO2 Humanoid shape Face Recognition Luckily I know the perfect sensor for you. Just go to adafruit. This is the best sensor I could find. http://www.adafruit.com/products/189 Good Luck. Form: Noah

fmilburn · Answer

Artificial intelligence senses people through walls | MIT News

rsc · Answer

Install electroencephalography sensors in their pillow(s) and match their brainwave patterns to your desired limits. Trigger on unconsciousness and send an email or text to the family.

Former Member · Answer

Hi Mrigank; A single sensor would be unlikely. A sensor array exists that can do this - It's called a polygraph, or "Lie Detector" (please forgive me if I sound patronizing, as that's not my intent). If you're not familiar with the device it simultaneously measures heart-rate, respiratory rate, galvanic skin response, and blood pressure. It's certainly possible to use existing MEMS technology to miniaturize to a single device. It doesn't measure chemical mediators (adrenaline, dopamine, etc.), so it couldn't be said to "measure emotions"; just their physiologic response. If you added skin perfusion (as in blushing), eye movement, and extremity tremor you could get a more accurate measure of what emotional response was taking place.

Former Member · Answer

You're welcome Mrigank! Copy on multiple people spread over an area. I'll try to avoid being verbose. This is a two-part problem that has been, and continues to be, worked on with varying degrees of success. The other factor you might be able to consider is using "choke points" so you can narrow down the subject pool. Most of the time these are turnstiles (to enter an area) or single-file lines. Even in an urban area, though, you can put up sidewalk barriers so people have to funnel through two or three at a time. That decreases your subject data points considerably. Part One: Isolating input to a single person. Facial recognition software and a high-resolution camera input has come quite far in recognizing a human face. Limiting factors are density of data (how close in proximity to each other the subjects are), interference modalities (hats, glasses), and certain facial angles (basically too much noise in the sensor input). For example; if you walk into a US airport with a cap and sunglasses on you're likely to be questioned, as the Computer Vision algorithms have difficulty performing facial recognition. Most of the actual devices and algorithms that do these kinds of observations are either classified as state secrets, or are at least highly guarded proprietary secrets. Part Two: Defining what is an unusual response, such as someone "panicking". Looking at airports, again, it's difficult to differentiate who might be nervous before they launch an attack vs. someone who doesn't like flying. The sensor analyses are usually done via computer algorithm. I can't go into specifics, but current open-source data points for emotional recognition are pupil response and skin temperature. Strictly speaking - No, there is not a sensor that detects human emotions from a distance. Emotional expressions are too variable to read externally. That's why people can lie to each other. Using currently available sensing devices combined with computer algorithms allows for a person to be flagged for questioning and analysis by a human.

shabaz · Answer

Agree.. Also I find it interesting how the title and question re-occurs regarding the word "sensor", when it really isn't a sensor issue as much as what can be done with the sensed data. The sensor on its own, and the data on its own is useless.

There's a reason why an accelerometer in a mobile phone, or in other hardware, is so useful. Not because we want to sense acceleration, but because with processing and that single sensor type, we can detect so many different situations (to certain levels of success).

I'd suggest it could even be possible (with suitable processing) to detect human presence reasonably well with even a magnetic sensor, because nearly everyone has something metallic, such as a belt buckle. It's success rate may be as good as any single simple sensor that can be fooled.

It's less to do with the sensors, and more to do with how well information can be extracted from the data (perhaps from one or more sensors) to achieve the task (in this case the task of identifying or classifying humans).. and to do that would require data capture, and lots of training of a model with known humans.

There was a time when processing was expensive, and all we could rely on was (say) an op-amp to help clean up the sensor data, and then we could make some interpretation from that data. So PIR sensors etc were simple. But adequate for a lot of uses.

The state-of-the-art is not a sensor+op amp any more. There's lots of processing that can be done at (or close to) the device/sensor level because computing is cheap(er).

dougw · Answer

I used to spend more on a thermocouple than the cost of a Raspberry Pi, way more if you include the meter. A Raspberry Pi with a PiCam can do a pretty good job of recognizing people.

dougw · Answer

I once developed a system where fingerprints were used for authentication. The users did not want their fingerprints captured, so we had to be rigorous about only using characteristics and never saving an image. It was not possible to recreate a fingerprint from the characteristics, but they could easily identify a person.

Former Member · Answer

Try the OMRON thermal sensor D6T available from Digikey and Mouser.

deeps · Answer

Hi This may be a 2 years late reply but you can use pedestrain detection in opencv Pedestrian Detection OpenCV - PyImageSearch today open-CV output has become much more robust and reliable and you can use it with any small board which has Linux and some camera interface. Also the more you increase camera quality the better the results will be.It is Cross platform and you can use c/c++/python/java/ruby and many more languages with same library Also I have a vivid Idea of having sensors beneath the roads and tiles which detect human like ( 2 legged) footsteps to determine presence of a human. A network of sensors beneath our ground that tracks not just our presence but our motion, direction and position. How about this way of doing it ?

Former Member · Answer

Listen I have solution like that you can use 2 ir sensors at your door and make a visitor counter . At each entry at your door, controller will count it, so we can say that there is person in home.

shabaz · Answer

Hi Gurinder, the OP specifically rules out PIR sensors: Am aware of PIR ( Passive Infra Red ) sensor , which would human motion but not the human presence . Anyway, I don't think a sensor exists to 'detect' humans, just some limited characteristics that humans have (and often these characteristics are not exclusive to humans either). Image recognition usually identifies anything that looks human-shaped. PIR detects anything warm that moves (cat or dog would be included). Also, plenty of 'human interaction' robots are designed to look and move like humans. They can speak too. Short of some intrusive surgery or extracting DNA, or interrogating and hoping that intelligent responses are present for only humans (and how long before this fails too?) I can't see how a machine or a sensor could 'detect' humans. Most of these characteristics can be faked too. This might sound pedantic but it's not any more, especially when cameras etc are used to unlock phones and PCs these days.

shabaz · Answer

Could you detect it? Without cutting up someone to confirm it actually is a human, and not a dummy? What if it was warm and produced a pulse, but was really a machine? If you could not detect it without intrusive surgery, how would you expect a sensor/machine to?

dougw · Answer

mrigank Check out the Omron system... https://www.element14.com/community/message/229828/l/looking-a-for-facial-recognition-hardware-tester#229828

Sensor to detect Human Presence

Top Replies