Is Microsoft Working on 'Performant Sound Recognition' AI Technologies? (windowsreport.com) 28
Windows Report speculates on what Microsoft may be working on next based on a recently-published patent for "performant sound recognition AI technologies" (dated April 2, 2024):
Microsoft's new technology can recognize different types of sounds, from doorbells to babies crying, or dogs barking, but not limited to them. It can also recognize sounds of coughing or breathing difficulties, or unusual noises, such as glass breaking. Most intriguing, it can recognize and monitor environmental sounds, and they can be further processed to let users know if a natural disaster is about to happen...
The neural network generates scores and probabilities for each type of sound event in each segment. This is like guessing what type of sound each segment is and how sure it is about the guess. After that, the system does some post-processing to smooth out the scores and probabilities and generate confidence values for each type of sound for different window sizes.
Ultimately, this technology can be used in various applications. In a smart home device, it can detect when someone breaks into the house, by recognizing the sound of glass shattering, or if a newborn is hungry, or distressed, by recognizing the sounds of baby crying. It can also be used in healthcare, to accurately detect lung or heart diseases, by recognizing heartbeat sounds, coughing, or breathing difficulties. But one of its most important applications would be to prevent casual users of upcoming natural disasters by recognizing and detecting sounds associated with them.
Thanks to Slashdot reader John Nautu for sharing the article.
The neural network generates scores and probabilities for each type of sound event in each segment. This is like guessing what type of sound each segment is and how sure it is about the guess. After that, the system does some post-processing to smooth out the scores and probabilities and generate confidence values for each type of sound for different window sizes.
Ultimately, this technology can be used in various applications. In a smart home device, it can detect when someone breaks into the house, by recognizing the sound of glass shattering, or if a newborn is hungry, or distressed, by recognizing the sounds of baby crying. It can also be used in healthcare, to accurately detect lung or heart diseases, by recognizing heartbeat sounds, coughing, or breathing difficulties. But one of its most important applications would be to prevent casual users of upcoming natural disasters by recognizing and detecting sounds associated with them.
Thanks to Slashdot reader John Nautu for sharing the article.
Glass break sensors are already a thing (Score:1)
Re: (Score:2)
Yes acoustic glass break sensors just listen for specific frequencies characteristic for breaking glass. It's very low-tech, literally just a microphone tuned to a specific frequency tripping a latch. You can get a glass break sensor module for under $10.
However, I'm assuming they're trying to figure out how to get a hugely expensive AI inference engine to run 24/7 on an audio stream and not cost an arm and a leg. Maybe they'll reinvent the wheel and just use a microphone with tuned to specific frequencies
Re: (Score:3)
Yes and they can be easily fooled. There's a store I frequent where they have such a sensor. They've triggered it a few times.
How?
The front security gate is a sliding gate with a locking bolt that goes into a hole in the ground and a bolt going into a hole in the track above. If you unlock the upper bolt without holding it, the noise of it crashing down sets it off.
If you slam the sliding gate shut with a bit too much enthusiasm it too will set it off. Or, since the alarm is usually counting down, if you op
Prevent casual users what what? (Score:1)
My 8 year old writes better.
Re: Prevent casual users what what? (Score:2)
Re: Prevent casual users what what? (Score:1)
You got to my comment so apparently you didn't :P
Well, it is Microsoft (Score:1, Troll)
So whatever it is, it will suck, be unreliable, be insecure and generally badly thought out.
Re: (Score:2)
They are trying to find a use for all that idle Azure server capacity.
Re: (Score:2)
But it will be "proformant", that darling word of marketeers everywhere.
Q: Why is your whizzy thingy better than this other one?
Marketeer: Ours is Proformant!!!
Q: Errrr....I give up, what does that mean with respect this other one.
Marketeer: Theirs is not Proformant!
Q: Why isn't theirs proformant?
Marketeer: The word is "Proformant!", and theirs lack Proformicity.
Q: I'll try again, why is your proformant.
Marketeer: Listen the word is "Proformant!". Ours is proformant because....because...(rustles notes)....t
Re: (Score:2)
They are seriously using that? People must be even more gullible than I thought...
Re: Well, it is Microsoft (Score:3)
Can it be used for Evil? (Score:2)
Microsoft Behind the Curve (Score:5, Insightful)
This is already built into the Amazon Echo, since 2019. It can detect glass breaking or smoke alarms and then alert you via push notification. You now have to pay for the feature (see "enshittification ") but it's hardly new. I guess if you slap the word "AI" on any old technology it can generate headlines and reassure shareholders.
Re: (Score:2)
What they could really do to make theirs stand out against Amazon's is make it recognize the sound of peeing dog.
Re: (Score:2)
I guess if you slap the word "AI" on any old technology it can generate headlines and reassure shareholders.
AI is a process not an outcome. Unless Amazon achieved something via a trained neural network they did something different. And yes it absolutely would reassure shareholders. If I were a shareholder of a tech company (or any company) right now, I would be asking questions why we're sitting around with our thumbs in our asses while the competitors are using ${today'shottrend} to solve ${problem}
Re: (Score:2)
Pretty weak effort if that's the goal, to be honest...
Using AI to solve already-solved problems is pretty low-hanging fruit.
Gedda real job! (Score:1)
Just. Make. Teams. Not. Suck. Stop playing with AI!
Re: (Score:2)
I'm not sure that this is even "AI"... it's just a lookup table of sounds.
Re: (Score:1)
The US Patent Office doesn't know the fucking difference if "decorated" with enough buzzwords.
Something for a AI to do (Score:2)
What would be nice: A microphone that takes the outside sounds, filters out the traffic noise, and plays back the bird song at normal level as if I was outside. Plus a UI that allows me to select the one barking dog that gets on my nerves, and seagulls.
I think Samsung makes a TV for windowless kitchens with a camera that can show the outside world - as if you had a 50 inch window.
I'm sorry. I'm weak. I really tried not to do this (Score:2)
I hear the first, least challenging test for Microsoft's Performant Sound Recognition AI Technologies will be to identify sounds associated with Fox News hosts and Republican Party power brokers fellating Donald Trump. If it manages to learn this task, this AI will be qualified to let users know a political disaster is about to occur.
You have a new notification: (Score:1)
Windows without blue screens (Score:2)
You can patent it but that doesn't mean you have to prove it to the patent office.
How about the sound of (Score:2)
Just collecting copyrights right? (Score:1)
It is nothing more than building a human verified training set of image to what's it a picture of, audio to what's that sound from, and text to 'what's the summary?'
of just about everything in existence.
https://www.medicaldesignandou... [medicaldes...urcing.com]
Medical teams using highly sensitive microphones to record the body to build a sound to disease mapping.
it's "performant", sheesh (Score:2)