Google’s Song Recognition Technology: Understanding the Ease and Accuracy of Music Identification
Instant music identification is now easier than ever with the help of modern technology, and for many people, that means using Google’s voice search: “Hey Google, what song is this?”. But what actually happens when we put these words together? How does Google know what song users are looking for and how could it possibly be so fast and accurate?
The Technology Behind Music Identification
Any time a user asks Google to “what song is this?”, they’re triggering a technology known as a “fingerprinting algorithm”. This is a form of artificial intelligence that uses audio features or “fingerprints” to determine what a sound or song is. What’s unique about fingerprinting technology is that it is able to work with small snippets of a song. When it receives a query, such as “Hey Google, what song is this?”, it will search its database of audio fingerprints for any matches.
The accuracy of the results can sometimes depend on how clearly the query can be heard, as it’s more likely to recognize a song that has been recorded in high quality. That’s why if the snippet of sound you’re asking about is noisy or muffled, the audio technology might not match it correctly.
Data and Multi-faceted Solutions
To deliver more accurate results to users, companies such as Google have to have a huge database of audio fingerprints, which they have been able to gradually build up thanks to user requests, or manual audio input.
In addition to Google’s audio recognition technology, they now also use other information when responding to ‘Hey Google, what song is this?’ requests. This includes contextual clues such as the time, the location, and even other sound sources that can help Google better understand what the user is looking for.
One example of this is if the user is at a karaoke bar and they ask “Hey Google, what song is this?”. If the words of the song can be heard clearly enough, and if there is any other contextual information from the surrounding environment, then it is likely that Google will be able to identify the song with higher accuracy.
Google also now uses an internal library of popular songs, so when a user searches for a popular song, it is more likely that the song will be found more quickly.
Final Thoughts
Overall, it’s incredible how quickly and accurately Google’s audio recognition technology can identify songs. While it’s not 100% accurate, it does get better with each passing year, as Google continues to use AI, data and contextual clues to make its results as comprehensive as possible.
So the next time you find yourself curious to know what song is playing, don’t hesitate to ask Google: “Hey Google, what song is this?”. It’s the perfect way to find out the answer to your musical mystery in no time at all.