Ever had a tune stuck in your head but couldn’t quite place the name? It’s a common frustration for music lovers. Imagine being able to simply hum a few notes and instantly find out what’s this song. Thanks to advancements in machine learning, this is now a reality. Think of a song’s melody as its unique fingerprint – each one possesses a distinct identity. We’ve developed sophisticated machine learning models that can effectively match your hum, whistle, or singing to the correct “fingerprint,” bridging the gap between a fleeting melody in your mind and the song you’re searching for.
So, how does this technology actually work when you’re trying to figure out what’s this song? When you hum a melody into a search engine, our machine learning models spring into action, transforming the audio you provide into a number-based sequence. This sequence acts as a digital representation of the song’s melody. These models are meticulously trained to recognize songs from a diverse range of audio inputs, including people singing, whistling, or humming, and even from studio recordings themselves. Crucially, the algorithms are designed to filter out extraneous details like accompanying instruments, the unique timbre of a voice, and variations in tone. By stripping away these layers, what remains is the core essence of the song – its number-based sequence, or as we call it, its melodic fingerprint.
The next step involves comparing this melodic fingerprint against a vast library of thousands of songs from across the globe. This comparison happens in real-time, rapidly identifying potential matches for what’s this song might be. Consider a popular song like Tones and I’s “Dance Monkey.” You can instantly recognize it whether you hear the studio version, or someone singing, whistling, or even just humming the tune. Similarly, our machine learning models are trained to recognize the underlying melody of the studio-recorded version. This capability allows the system to accurately match it with the audio of a person humming that same tune, effectively answering the question of what’s this song.
This innovative feature is built upon the foundation of our Research team’s prior work in music recognition technology. We initially introduced Now Playing on the Pixel 2 in 2017, leveraging deep neural networks to bring low-power music recognition directly to mobile devices. In 2018, we extended this same technology to the SoundSearch feature within the Google app, significantly expanding its reach to a catalog encompassing millions of songs. This new humming-to-search experience represents a significant leap forward. Now, identifying what’s this song is possible even without relying on lyrics or the original instrumental track. All it takes is a simple hum.
