Taking audio pattern recognition to the next level…
Pop Music
One of my favorite “Big Data” presentations is the one from Shazam at the O’Reilly Strata + Hadoop World 2015 by Cait O’Riordan, where she shows us how Shazam can tell if a pop song is going to be a worldwide hit from usage patterns of the 100 million plus users of the Shazam App, three months before the pop song will actually become a worldwide hit.
Now from some ideas and predictions as we look into the acceleration of technology and software in this audio fingerprinting space.
First, let’s stay in the realm of music and musicians. With advanced audio fingerprinting technology it would soon be possible to discern and identify individual vocal artists from a track. Which means in the future when you use you pop song identifier it will be able to give you the name of all the background singers, it will also be able to recognize each individual musical instrument and piece of technology that was used to lay down the track. Beyond that, after identifying each musical instrument it will be possible to match the audio fingerprint for s specific player on the instrument so your app will be able to tell you who played the sax on the track.
Then there is also the matter of recognizing the specific recording technology that was used and what medium, perhaps even one day it can recognize the acoustics of a specific recording studio as in this is the audio fingerprint of this London recording studio in 1969.
Let’s move on from music and song to other more annoying and less entertaining forms of human-produced noise.
Planes, Trains & Automobiles
With 1.015 billion motor cars on the road in the world today the next “killer app” might very well be a “Shazam for cars”. Besides, being able to identify the make and model from the internal combustion engine noise the most obvious commercial application would be to tell you what the state of your car’s engine is. Does it need an oil change? Is the engine timing out? Finally, you will have the ability to identify “That Funny Noise” that your car makes and what to do with it. Now we can extrapolate that to all other noise-making engines as well.
Guns
Across America, cities are silently adopting the ShotSpotter a network of microphones that will identify and triangulate a gunshot. Let’s look at the future possibilities here. A next level shot spotter will be able to identify the make and the caliber of the gun being fired, beyond that it will be able to calculate all possible bullet trajectories in real time and listen for and identify the bullet impact from which it can narrow down the exact trajectory of the bullet.
A possible military application of this is a network of smart microphones that can identify and track all forms of incoming ordnance fire, soldiers could be outfitted with HUD helmets that will show the real-time heat maps and tracks of enemy fire. Going from using audio fingerprinting tech for military defense it could also be used for offensive technology. Heat seeking missiles can be made smarter by transposing the heat signal with the audio signal of an enemy aircraft for better accuracy, beyond that it would be able to identify the make and model of the plane and discern whether it is locked on an enemy aircraft, or if presented with multiple targets go for the highest value target.
Where are we now?
Open source music fingerprinting
Vehicle Sound Signature Recognition by Frequency Vector Principal Component Analysis