How often does it happen to you that a song gets stuck in your head and you keep humming it but to no avail? It happens with me all the time.
We all know how irritating it becomes when we can’t remember anything about the song but its melody. We ask our friends and everyone around us to help us identify that song.
Well, Google has come up with a solution that helps you identify the song that keeps playing in your mind. Let’s know about Google’s new ‘hum-to-search’ feature that allows you to find your latest earworm.
What is it?
Recently, Google added a new search feature which lets you hum, whistle or sing a melody to identify the song that is stuck in your head. Don’t worry about pitching or humming perfectly. This feature works well even for all those bathroom singers as well. Google uses the machine learning algorithm and displays the list of potential song matches.
To use Google’s ‘hum-to-search’ feature, all you need to do is open Google app or Google Search widget, tap the Microphone icon and say “what’s this song” or tap the Search a Song button. Then you can hum, sing or whistle the tune of the song you are looking for. Google will run its algorithms and display the list of songs that match with the melody.
In case you’re using Google Assistant, say “Hey Google, what’s this song?” and then start humming the tune. Google will display the results instantly.
Google has rolled out the new feature worldwide for both Android and iPhone users. The new feature is available in English for iPhone users and in more than 20 languages for Android. However, this feature works on mobile devices only, and not on your Google Home or Nest speakers.
How does it work?
For all the curious heads out there, this is for you. Google uses machine learning techniques to find a song that is stuck in your head. Let us understand how it works.
Each song has a unique identity, and its tune is its fingerprint. Google has built machine learning models that transform the melody that you hum into a number-based sequence. The machine learning models identify the songs based on humans singing, whistling or humming, and studio recordings. The algorithms also take into consideration other details, like accompanying instruments and the voice’s timbre and tone.
Google then compares these sequences to thousands of songs from around the world and identifies potential song matches in real-time. The machine learning models recognize the melody of the studio-recorded version of the song and use it to match with what you hummed.
Google says that this new feature builds on the work of its AI Research team’s music recognition technology. In 2018, Google used deep neural networks to the SoundSearch feature in the Google app and expanded it to millions of songs. The ‘hum-to-search’ feature is Google’s step further because now it can recognize songs without the lyrics by just a hum.