[urban_winter]

This is just a dedicated RF emitter combined with a dedicated receiver. The fact that is it uses WiFi hardware is probably just because that's the cheapest and most available hardware for the researcher to work with. There is no indication in the article that the WiFi can actually be used for transmitting real data at the same time; that a non-dedicated WiFi source can be used; that it works when there are many people between transmitter and receiver.

Therefore the ideas that this might apply to real-world situations and use existing WiFi infrastructure, are a stretch given the information that's been shared.

It basically doesn't seem like a big deal to demonstrate what has been demonstrated.

[thomascountz]

Research doesn't always seem like a big deal. In this case, using CSI extracted from standard wifi packets (beacons, data frames, etc.) from commodity hardware is the core of the "big deal."

In principle, any packet that carries data can also be used for sensing, though, as you mentioned, this isn't what the researchers demonstrated. However, for years, this kind of thing was studied using special multi-antenna Intel cards to get a clean signal. Getting this level of accuracy from such a low amplitude signal from a single antenna on commodity hardware like an ESP32, is the actual breakthrough. It proves the concept is sound before tackling the much harder problem of using a standard home router amidst other traffic or isolating multiple targets in a room.

[AyyEye]

Mmwave heartbeat sensors are like $2 at retail pricing. This is commodity stuff, I fail to see how adapting it to a new radio is any kind of big deal.

[thomascountz]

Using channel state information is not about "a new radio." CSI is a byproduct of existing WIFI standards/infrastructure.

CSI does require a supported chipset, like an ESP-32. However, if an IoT device is already using an ESP-32, for example, one would not need to add dedicated hardware (like an mmWave MR60BHAX) to be able to do things like presence, breathing, heart rate, and location detection.

As a hobbiest/ESPHome user, I have lots of ESP-32s and not lots of mmWave-s. As a business, I'm already shipping with an ESP-32 and I don't want to increase my BOM.

Besides this, I find this research to be a big deal as it has implications for privacy and security. Your biometrics can be collected using existing widely-deployed hardware using existing internet standards. Your smart toaster can indeed be spying on you in more ways than you think.

But anyway, using CSI for sensing will soon be old hat. IEEE has granted approval to the 802.11bf WLAN Sensing working group to define standards for exactly these types of applications. Taking what's currently an artifact of an implementation detail, and turning into a first-class feature.

Edit below

I want to point out another thing: "clinical-level heart rate monitoring with ultra low-cost WiFi devices" can be lifesaving in situations where clinical-level heart rate monitoring is otherwise unattainable.

[amelius]

Human vital signs detection can be useful in earthquake situations. Measuring through walls of concrete is difficult, however, so a new radio is needed (lower fequency and/or more sensitivity).

[boxed]

> This is just a dedicated RF emitter combined with a dedicated receiver. The fact that is it uses WiFi hardware is probably just because that's the cheapest and most available hardware for the researcher to work with.

Ok.

> There is no indication in the article that the WiFi can actually be used for transmitting real data at the same time

So? No one said it was.

> Therefore the ideas that this might apply to real-world situations and use existing WiFi infrastructure, are a stretch given the information that's been shared.

What? First you say it's trivial/obvious, and now it's impossible? Decide on your critique.

[urban_winter]

The dominant themes in the thread relate to using existing WiFi infrastructure in real world environments. I thought it would be obvious that I was critiquing this line of thinking. Obviously not.

[transpute]

Real-world applications benefit from recent on-device hardware like NPU or Apple Neural Engine.

Intel demo on commercial laptop (2022), https://news.ycombinator.com/item?id=45130061

Qualcomm human-in-home positioning demo (2021), https://www.youtube.com/watch?v=xNmnqCsvMTU