The US has developed a drone detection system that converts sound into a 3D image.

Specialists from Hall Lidar (USA) have created the UDL-64 UAV detection system, which converts sound into a 3D image. It incorporates artificial intelligence to provide complete situational awareness of the surrounding airspace.

This was reported in the press service of the company.

They explained that the system allows for the formation of a spatial representation of activity drones based on audio information.

The design is lightweight and compact. The device's diameter is approximately 60 centimeters. This allows for quick and easy deployment. It can be used to monitor air traffic at both fixed sites and mobile military units. The system can be mounted on a tripod or mast and deployed in minutes.

The device features built-in Wi-Fi, Bluetooth, and 5G modules, simplifying its connection to a wide range of control and monitoring systems.

The UDL-64 operates without the use of radio frequency (RF) or GPS. This allows it to operate in passive mode and remain undetectable to enemy sensors, as it emits no signals.

In autonomous mode, the device can detect drones at a distance of up to 200 meters. If several are deployed simultaneously, the detection zone expands to approximately half a kilometer.

Comments

[Mini Mokik]

Shazam at maximum speed 🥴

[poquello]

Quote: Mini Mokik

Shazam at maximum speed 🥴

In 2025, seventh-grader Mikhail Lyapkin created an acoustic-optical system for recognizing UAVs in difficult terrain. The device can detect drone sound at a range of 300–400 meters, and up to 2 kilometers with a more powerful microphone. Recognition accuracy was 96% after testing.

[Mini Mokik]

After Apple bought Shazam, he can recognize any music from 20 kilometers away 😏

[poquello]

Quote: Mini Mokik

After Apple bought Shazam, he recognizes any music from 20 kilometers away 😏

Have you heard anything about sound wave attenuation?

[Mini Mokik]

So? What difference does that make if everything in the US is Amazing and nothing else? 😑

[poquello]

Quote: Mini Mokik

In the USA everything is Amazing and nothing else?

Maybe Magic would be more suitable? Everything is so fabulous,
How many decibels are needed for 20 km?

[Mini Mokik]

You just need to talk about it for 10 seconds. No more 🤔

[poquello]

Quote: Mini Mokik

You just need to talk about it for 10 seconds. No more 🤔

Alice on Android does the same thing.

[Mini Mokik]

An old man comes to the doctor:
"Doctor, I need your help. I'm having problems with my potency. I can only see my grandmother three times a week."
“And how old are you?” asks the doctor.
- 85.
“So you, my dear fellow, are a sexual giant,” the doctor is amazed.
— But doctor, my neighbor is also 85, and he says that he and his grandmother have sex every day.
- My dear, you tell us too!

[poquello]

Quote: Mini Mokik

- My dear, you tell us too!

"Advertising: Create a story that speaks for you."

[Green tea]

Does it detect the presence of a UAV? Or does it accurately indicate its location in three coordinates?

[poquello]

Quote: Green Tea

Does it detect the presence of a UAV? Or does it accurately indicate its location in three coordinates?

wow, "Bring me Lyapkin-Tyapkin!"

[Green tea]

Playing the fool instead of having a constructive conversation is typical when you have nothing else to say, but you still want to. The seventh-grader is great, but the Yankees rolled out a working prototype, essentially a system for automatically aiming cannon artillery.

[poquello]

Quote: Green Tea

to act the fool instead of having a constructive conversation,

God be with you, look for it yourself ))))))))))))))))))))))))))),
when you don't feel like searching, but you want to

[Hitriy Zhuk]

It is NOT a problem to attach a microphone and camera to a laptop.
But to put it into production...

[Irokez]

I can guess how it works. It's similar to a video camera's sensor, but instead of light-sensitive elements, it uses sensitive directional microphones.

[alexboguslavski]

The basics have long been known. The rest is just a matter of technique.

[Deaf]

And also some kind of AI. Pixel shaders are also AI then. Just like DLSS, FSR. Any program is no smarter than its creator. We used to be able to use Nano consoles like they were the apple of our eye. Now, it's like, "Oh, it's AI."

[Irokez]

Quote: Deaf

Now - oh, like it's AI.

Everyone's so scared of AI. But AI is an Intelligence Simulator, and some people already think that automated systems that rely on relays and all sorts of sensors and timers are AI.

[Hitriy Zhuk]

Quote: Deaf

Just like DLSS

You still won’t believe it!
That's it.
A system trained to do certain crap.
It's just that AI is a general and commercialized nickname for a bunch of methods, techniques, and mechanisms.
It's time to punch him in the face for this.

[Deaf]

Quote: Hitriy Zhuk

It's time to punch him in the face for this.

I agree. I'm currently working at KVI. We actively use our systems. Well, they're also called AI.
But I've been messing around with the code. On my home computer, I use it to analyze Telegram messages, for example. I've tweaked some things myself for specific purposes. It's pretty good at identifying characters based on phonetics, writing style, and so on. The algorithms, I think, are clear. It's just an add-on to big data—nothing more, and it's not AI. There's a separate NPU. It's an AMD, though a first-generation one. It certainly eats up a lot of memory. I'm already considering buying 2x48 6400 DDR5.
PS: DLSS 4.5, by the way, actively uses the NPU. Although there's a glitch with the load display in Win11—it doesn't show it at all. But you can feel the laptop getting warm. Even though the GPU load drops to 70-80% (4060RTX discrete graphics).

[Michael]

Quote: alexboguslavski

And then - a matter of technology.

Incidentally, it wasn't just technicians. During the siege of Leningrad, the best sound equipment teams were made up of visually impaired people. Blind people often have very sensitive hearing.
An experienced sound-detecting system could determine the type of aircraft, their number, their direction very precisely, and their altitude quite accurately.

[Deaf]

I have a visual impairment due to a concussion. -8. I haven't worn contact lenses for 25 years. I even fought a bit while disabled. My hearing is excellent. I think you know the "click" and "horse" exercises. Back when cans were plentiful, I could tell the range and azimuth of a shot by ear.

[Michael]

Quote: Deaf

I think you know the "click" and "horse" exercises.

I can guess, but please enlighten me. I'm more familiar with the "turtle" exercise.
And without any undue sympathy (I am disabled myself), I will say one thing: A man is when you have balls, and those that are in your head and not made of plasticine.

[credo]

Specialists from Hall Lidar (USA) have created the UDL-64 UAV detection system, which converts sound into a 3D image. It incorporates artificial intelligence to provide complete situational awareness of the surrounding airspace.

If it also works just as successfully against earthly targets - humans, machines, or animals - then its range of use becomes significantly wider.

[Kmet]

It's unlikely to affect humans or machines, as their sound profile differs significantly from that of UAV "cutting blades." It closely resembles the sound profile in water (when a diver dives or surfaces) near a passing vessel (boat, speedboat, motorboat)—a distinctive sound that transitions into a chopping whistle, like that of a shredder, is clearly audible (although, of course, it's not the same frequency range). The UAV sound profile is primarily composed of a mixture of the low-frequency range transitioning into the midrange (150-400 Hz), plus a "picture" of the upper midrange dropoff and a shift into the lower HF range (the whistle/sharpness of a chopping knife: 800 Hz - 3,5 kHz—upper midrange and lower HF) plus a slight squeak at 4 kHz (HF). The human friend's "image"—the energy of the voice frequency range—lies in a narrow band of 300-3400 Hz, while footsteps (short impact noise with rapid attenuation) fall in the 50-500 Hz spectrum. Cars, on the other hand, create a "image" of low-frequency noise in the 30-150 Hz range (engine hum) plus broadband tire noise with a peak at 500-1200 Hz. Clearly, this device is tuned to a very specific frequency range (sound image), which must filter out all others for clear identification (the absence of false alarms). The drone creates a continuous, modulated whistling-buzzing background, while a person or machine produces pulsed or smoother sounds. In general, this particular model (I am more than sure) is precisely tuned to its "sound image", but in general, the technical implementation can provide for work with several identification ranges (although with the sensitivity of the low-frequency and infrasound sensor, there will be certain difficulties in setting up and filtering/identifying in terms of "parasitic interference").

[credo]

It's unlikely that it would affect a person or a machine, as their sound profile differs significantly from the sound profile of the "cutting blades" of a UAV...

It's a pity if it's as you describe.
On the other hand, if this is true, then the effect of detecting a UAV at the specified distance becomes insignificant if the UAV hovers above the target at a significant altitude and drops a payload. This makes the system limited in scope and highly specialized.

[Kmet]

You've reached absolutely the right conclusions. Frankly, the claimed 200 meters isn't much of an identification range, considering the speed of drones. Consider the fairly standard speed of a drone—120 km/h—that's 33 m/s, meaning it can cover those 200 meters in about 6 seconds. But what if the drone is faster? That's the point of such a thing, even if you were to combine them into a whole "field." Maybe I'm missing something or misunderstanding something, but in a military context, it seems to me like a device of questionable "usefulness."

[Askold65]

Quote: Kmet

Let's take a fairly standard drone speed: 120 km/h (75 mph) is 33 m/s (33 m/s), meaning it can cover those 200 meters in about 6 seconds. But what if the drone's speed is higher? That's the point of such a thing.

The idea is to place it several kilometers from the protected object in order to send an "Alarm!" signal from such and such a direction.

[Kmet]

In theory, I agree, but what about practical application? Let's approach the question objectively: how many such detectors would need to be installed, over what area, would drones destroy them, and who would maintain all this equipment (install/dismantle/upgrade)? The idea is good until it's put to practical use. It's one thing to install them on a stationary object (as proposed – at bases), but another thing entirely to deploy them in combat on mobile ground stations. That's where the questions arise. More precisely, doubts about their applicability.

[Askold65]

Quote: Kmet

How many such detectors need to be installed?

These calculations should be done by specialists.

Quote: Kmet

won't the same drones destroy them?

How so, if these devices operate in passive mode and are practically undetectable?

Quote: Kmet

It is one thing to install it on a stationary object (as proposed - on bases), but another thing is combat use on a mobile combat station.

Absolutely right - around the bases at a sufficient distance from them to provide timely warning of approaching danger.

[Kmet]

More detailed technical specifications for the Hall Lidar UDL-64 system were only found on the following resources: https://nextgendefense.com/us-ultimate-drone-listener/ or https://i.gasgoo.com/news/70453940.html, as the official page (https://halllidar.com/) does not provide detailed information about the device.
What they write, among other things:
Integrated connectivity features such as Wi-Fi, Bluetooth, and 5G allow the system to be easily integrated into a wider range of command and control architectures. As a standalone device, a single UDL-64 can detect drones at a range of up to 200 meters. When combined, multiple units can extend the coverage area to approximately 500 meters.
Given these detection range data, specialists likely won't be able to help here, as everything is limited by the communications technologies used (Wi-Fi, Bluetooth, and 5G) and network architecture. Therefore, it's definitely not suitable for LBS (electronic warfare will overwhelm this network), and there are also limitations for stationary objects (you can, of course, place detectors a couple of kilometers from the object, but the detector field's range is limited to 500 meters—meaning that beyond that, you can only rely on luck and the UAV's "approximate" direction. If you use a system with a 500-meter radius directly in front of the object, then taking into account the time it takes to accurately detect the drone, transmit an alert, and react to the threat (another 5-7 seconds for processing)... considering that the drone isn't stationary, but is moving toward the object, for example, at 60 km/h, the drone will cover these 500 meters in 30 seconds of pure flight time). And this doesn’t even take into account false alarms, detection duration, the drone’s flight altitude, and its maneuvering when approaching the target.
Passive detection is out of the question (given the device's updated specifications), as the drone will prioritize the target itself, not the detection system. If it's a flight mission, it will definitely ignore the detectors; if it's a PPV, the operator can see any target and prioritize the one.
And also... if you take into account how they jam our GSM communications and the internet during attacks, then everyone's chances are proportionally reduced (one to fly, and the other to detect and transmit data).
This system is ideal for a quiet acoustic background, which is a rarity even for oil refineries (not to mention LBS, where the system will simply constantly lie and will have to be coarser).
In general, this is applicable, in my opinion, only to a comprehensive air defense system, and even then the question of the efficiency and effectiveness of this particular system becomes highly controversial.

[your1970]

Quote: Kmet

Let's take an objective approach to the question: how many of these detectors need to be installed, over what area, won't the drones destroy them, and who will maintain all this "equipment" (install/dismantle/complete)?

Oil refinery/thermal power plant/nuclear power plant/and a mountain of other facilities -
Early warning system

[Kmet]

For fixed installations around a protected facility, yes, but again, the question remains: is the coverage altitude of these detectors (the sound detector's conical horn points upward) also 200 meters? This altitude issue has already been discussed: a drone can perform drops or dives from high altitudes. Let's assume a drone approaches the detector's zone at an altitude of 300 meters, selects a target, and attacks. In theory, that leaves 6-12 seconds for a reaction (a near-vertical dive). What can the defending facility do during that time? Most likely, they'll only be able to detect the attack's origin, leaving no time to repel it. If we're talking about a drop, the situation is even more dire, as the drop may not be identified. In general, it would be interesting to see the actual radiation pattern of the detector—that would eliminate a lot of speculation.

[your1970]

Quote: Kmet

This altitude issue has already been discussed: a drone can perform drops or dives from high altitudes. Let's assume the drone approaches the sensor zone at an altitude of 300 meters, selects a target, and attacks.

А above it will be detected by conventional air defense
Here is the good old problem of "300 meters" of infantry

[Kmet]

This again depends on the drone's size. A Mavic or light FPV is very difficult for air defense systems to detect due to its size, and there aren't as many Pantsyrs as we'd like. Something like a Geranium—yes, the radar will clearly detect it, but only after taking into account maneuvering, flight speed, and detection time. How much time does an air defense system need to detect and destroy such a target? And how much time will it have to spare to avoid hitting its own target, as the Patriots did? But for a light combat system, such a system is too ambiguous a solution, as it will produce a lot of errors (it's noisy), and recognizing drones using the "friend or foe" system is impossible. Another advantage is the UDL-64's proprietary communications and data transmission system—a separate issue.

[your1970]

Quote: Kmet

It's very difficult for a Mavic or light FPV to distinguish from air defense systems due to their size, and there aren't as many Pantsyrs as we'd like. Something like Geraniums—yes, radar will clearly detect them.

It is precisely against long-range missiles at stationary object sites in the depths - on the LBS, naturally, against the Mavics it is not a masterpiece (the last line of defense).
Although, let's say, if the lines are 20 km apart, it makes sense.

[Kmet]

Frankly, in my opinion, such a system should only work as part of a comprehensive system, not as a standalone solution, as the likelihood of false alarms or missed targets is very high. Overall, what can I say is that they're trying to find more effective solutions for UAV detection, and trial and error will eventually lead to something.

[Kmet]

The whole issue isn't just about the alarm response time, but also about detecting the target, which must be detected before it can strike. The human factor here is minimal and insignificant (due to the time factor), but for an automated anti-aircraft mount, it's perfect, but with some caveats (the target still needs to be found and locked on in the air). Furthermore, there are nuances regarding the altitude of UAV detection, as a system of such detectors has a flaw not in horizontal detection, but in vertical target detection. Flying a drone at an altitude of over 200 meters minimizes the system's advantages.

[stankow]

And in terms of protection of oil refineries?

[Kmet]

That's true, but why are we only considering horizontal attacks? I agree with horizontal attacks (when the drone is prowling and searching for targets), but what if the attack is on a stationary object, the target's coordinates are known, and there's a clear flight mission? A UAV can approach a target from an altitude of over 200 meters, and the entire detector field would be useless. So, it turns out that such an object would be detected (presumably) from a distance of 200 meters (judging by the device's specifications). So what are the strengths of the proposed system? Moreover, UAV sizes can vary, and not all radars can reliably detect them yet. The goal of this system is to ensure early threat detection. That's precisely why there's skepticism.

[Askold65]

Quote: credo

If it also works just as successfully against earthly targets - humans, machines, or animals - then its range of use becomes significantly wider.

I remember back during the Second Chechen War, they showed a sound-detecting device on television at a Russian troop position located on the perimeter of the Chechen Republic's administrative border. Personnel claimed that human footsteps and conversations could be heard through it from hundreds of meters away.

[SAG]

can detect drones at a distance of up to 200 meters.

This gives a full 4 seconds to take cover for people who are hard of hearing 👏
I hope there will at least be a column with Alice?

[alexboguslavski]

Should we ban motorcycles? Does the Constitution allow it?

[Andriuha077]

The system contains 64 highly sensitive MEMS microphones arranged in a phased array antenna on a 61 cm diameter disk platform.

It is based on technologies that David Hall (founder of Velodyne Lidar) previously used in lidars, but adapted for sound.
Onboard algorithms perform "beamforming"—this allows the system to programmatically "direct" the microphones' attention to a specific point in space, cutting out noise from other directions.

A network of several UDL-64s creates an "acoustic dome." Triangulation allows for the precise determination of not only the direction but also the height of an object.

As new drone models emerge constantly, the system's software is updated over the air, adding new signatures to the AI ​​database.

Integrating the UDL-64 with optical cameras transforms the system from a simple detector into a fully-fledged target management system. The acoustic array acts as the "eyes and ears" that guide the camera to the target for visual confirmation.
In fixed configurations, multiple UDL-64 modules can operate with a single powerful camera.
The stealthy UDL-64 discs are installed along the perimeter of a fence or on a roof. As soon as the system detects a drone, it automatically deploys a hidden camera.

Hall Lidar also promotes Cam-Lidar technology, which combines lidar and a camera at 60 frames per second for high-precision 3D tracking of high-speed targets.

[alexboguslavski]

How will this system react if the first wave includes drones without jammers mixed with attack drones?

[Michael]

Quote: alexboguslavski

How will this system react if the first wave includes drones without jammers mixed with attack drones?

And with speakers broadcasting the "March of the Union" from Red Arlet.

[K-50]

The US has developed a drone detection system that converts sound into a 3D image.

It's buulshit.
It has been observed, and even since the Second World War, that animals, dogs and cats, caught in an airstrike, or as we have now, a drone strike, hear their approach well in advance; no electronics can achieve such a result.
All that remains is to organize the training.
It's a shame about the cats, of course, and it would be difficult to train them; Kuklachev is unlikely to take it on. But dogs are a problem. We have plenty of dog trainers and kennel clubs.
So, it is necessary to organize training of dogs for early warning about the threat of a UAV attack.

[Radius]

Good idea. Training isn't difficult. Any "noble" mutt can handle this kind of work.

[cpls22]

Passive systems are the future. Currently, she can hear at a range of 200 meters, but later she will be taught to distinguish distinctive sounds from noise at greater distances. Digital hearing will be added to digital vision.

[opuonmed]

Well, someone thinks about the future and creates compartment systems.

[Fangaro]

I tried to remember the math. What size are the cells on the surface? About 6 x 6 cm? And the receiver itself? 2 x 2?
I wonder what's inside.

[cpls22]

Quote from Fangaro

I wonder what's inside.

The key is likely software data processing. Similar to human hearing, the brain can isolate from a variety of noise only those signals it's focused on. If listening to a recording from a sensitive microphone, the average listener won't be able to discern anything without decoding, although if they were close to the microphone during the same recording, they wouldn't have any problems.

[cpls22]

The cells are most likely just a protective mesh. If there are only 64 microphones, then it's possible there's a system of horn-shaped honeycombs printed on a 3-D printer.

[Photon]

You can throw together as many systems as you need. And AI is completely unnecessary. A few good microphones, preferably with acoustic reflectors, and visualize however you want. The most well-known visualization system is matplotlib. But Makie is better, so you don't have to deal with the slow Python. And the seventh-grader did a great job of attaching the optics. AI wouldn't hurt either. But I haven't heard anything about the seventh-grader and his system.

[Hitriy Zhuk]

In theory, if you hang high, it won’t work.
The drone is not a jet, and not a Tu-95...