Hacker News new | past | comments | ask | show | jobs | submit login

No one is going to buy this. It doesn't jam your voice at all. It introduces enough distortion and noise so that computer speech recognition software gets 4/5 words wrong.

Let me make a comparison. Claim: Those capchas with "type in the word you see" have an error rate of 4/5 letters when read by a computer. Yet here we are, humans, reading them just fine.

Now, if you're a conspiracy theory fake elected moon landing virus hoax person who is afraid of the government having microphones everywhere, running a keyword search on microphones placed around the park.... Still not a jammer - I'm sure whatever voice recognition software this thing confuses, the government has much better ones. Simply from the fact that most of the people they're interested in have accents so thick it's worse than some added "jammer" noise.

But to answer your question, "sounds" like they're trying to go commercial soon, because their announcement of this dirt simple product sounds like a cheap advertisement for some dick enlarging pills. meanwhile, your phone speaker can make sonic sounds since they support a sample rate of over 44khz. So can your computer.

here you go. you can google and download some files to play too. https://mynoise.net/NoiseMachines/audioJammerNoiseGenerator....




I’m not worried about the government spying on my individually, the resources to fight that would be impossible

I am worried about large companies spying on me en mass (and selling or leaking that information). That requires automation (voice/face recognition), it’s also reality, not a conspiracy theory, alas I do see it as inevitable and there only so much you can do to fight it.


> Yet here we are, humans, reading them just fine.

I've met a lot of captchas that I couldn't reliably solve. I guess that proves I'm a robot.


Same here. My guess is that they constantly AB test/explore captchas to push the envelope of genuine humans users in order to make it as hard as possible for the bots.


This is great for meetings. Don't have to ask everyone to put their smartphone off. Just be in a private room without cameras (against lipsync and other private details leaking such as notes).

Re: Your computer/phone can make noise. They use an Arduino to power this thing, and it manages to block with natural body movement.

I'm pretty sure this is going to ruin ANC headsets too though.


So what is the effect in the first part of the video on the website then? Are you saying they‘ve faked it?


It just relies on cheap microphones being fairly non-linear.

A quality microphone, with a decent low-pass-filter would be immune to this kind of hack.

Think about it, why aren't our ears affected?


> why aren't our ears affected

they're affected listening to the raw microphone capture. that's because the noise generated is much louder than the speech. so listening to raw microphone capture with this thing on, you can't make out words at all - just a few syllables here and there. the key is that the full voice signal is still captured by the microphone. you then filter out the additional noise (your conference software and laptop microphone does that at a shitty level, rtx voice is the top consumer product for this). once the noise is gone, you have only the original voice. due to the filtering, the original voice is very lightly garbled, which is enough to confuse voice recognition software, not your ears.

notice how they're using voice recognition to measure success? that's misleading and on purpose. if they used a person to see if this product is effective, the result would be - the product is literally nothing but a loud noisemaker which in zero ways does what it says it does.

now, I don't have an rtx. if you want to screw around, here's a very shitty two megabyte software version of what rtx voice does, it's open source and just a couple of meg. https://antlionaudio.com/blogs/news/free-active-noise-suppre...

a low pass filter, or a nice microphone would not help you at all here - the white noise is in the vocal range and no simple frequency filter is going to remove it. their product makes ultrasonic noise, which (probably due to microphone resonance) makes the mic vibrate in the audible range. again, not affecting the original voice signal, so it's loud white noise overlaid on top of the voice.

it does seem putting something in front of the mic that absorbs ultrasonic frequency but not lower voice frequencies works well to block this. which is why the thin plastic waterproofing film over the phone microphone completely blocks this product - no noise removal filter needed.

So literally any smartphone with water resistance bypasses this thing and records fine. for the rest, just run a noise removal filter on the recording and listen to it. for governments, I'm sure their spy mics are waterproof and immune, and if not, their voice recognition is better than whatever consumer tech these guys tested it on.


> they're affected listening to the raw microphone capture.

Well, of course.

But I was talking about listening directly, not via a microphone.

The whole point of the device that the white noise is not evident to listeners in the vicinity. If it was, people would not be able to speak to each other.

Which means that it's caused by intermodulation in the crappy microphone. Therefore a better quality mike (and a LPF) would remove it.


No, I'm saying that does literally nothing but add noise. It does not block or change the original signal in any significant way for human ears - only for consumer-level voice recognition software. Which is extremely crappy even without noise for anything more complex than "hey google play some ads in the middle of this song."

Once you take the noise out, the voice recognition software has problems recognizing the words. People do not. You can take the noise out at home with lots of different software or hardware. Skype/teams/zoom do this on a very basic level by default - as do most microphone audio drivers in windows. Once it's taken out, distortion from the noise is just enough to make the voice recognition miss a bunch of words. Human ears do just fine though.

https://www.youtube.com/watch?v=u3zdGza5zeI


There is a short sample of audio in the YouTube link. Your video is not even remotely similar in sound. If you could take the audio from the original and extract easy (for humans) to understand speaking, your confidence that this won’t work would be obviously justified. But so far you don’t appear to have even understood how the device works.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: