SoundHound AI (SOUN -1.25%) boldly includes "AI" in its name, which naturally raises the question -- how much artificial intelligence (AI) is actually involved in its products and services?
Luckily, I can assure you that this AI claim is for real. SoundHound AI really does boast a unique and powerful collection of advanced AI tools.
How SoundHound AI's audio analysis works
You may have run across SoundHound's namesake music identification app on your smartphone. That's where the company started, as CEO and co-founder Keyvan Mohajer built and refined the Houndify audio analysis engine over the years.
SoundHound, also known as Midomi in the early days, used a different approach to other song identification tools. You needed a pristine recording of the original performance. Anything less would return an empty search result.
But Midomi, and then SoundHound, used a more advanced method.
First, the original melody is converted to several layers of hummed melodies. The lead melody gets one track, drum patterns get another artificial hum pattern, and so on. Then, you can hum or sing into your phone, allowing SoundHound's audio analysis process to match distinctive patterns of harmony and discord in the multi-track song database.
"It's simply easier to match a hum to another hum rather than match it to an original recording," Mohajer said in an MIT Technology Review article 17 years ago. Thus, a unique feature was born as no one else tried to analyze the user's own voice.
Yes, that's a serious AI solution
That's an advanced example of deep learning, and lessons learned back then still inform SoundHound AI's audio processing today. It's just a bigger business now with automakers, restaurants, and customer service call centers among its clients.
So if you run into an automated menu system on the phone or at a drive-through window, you may be talking to SoundHound AI's artificial intelligence systems.
