16 October 2020
Google held its “Search On” feature on October 15 where it announced some very interesting features around the search powered by AI. One of the top features is hum support on Google Search, which helps you find the melody that is stuck in your head. The feature uses machine learning to efficiently identify the song that you are just humming to it.
We often like a song that we heard somewhere in the past but we do not remember a single word from its lyrics. Take for instance you went to a shopping mall last week where a song was being played in the background but it is now stuck in your head. This song has now become no lesser than an earworm that is hard to clear off. Unfortunately, you only remember the melody of it and there seems no way to get to the right song. We hum or whistle to our friends but they also do not have any idea what we are looking for. Here’s when Google’s new "Hum to Search" feature comes to the rescue.
Google has announced that starting today, you can hum or whistle to Google to solve your earworm. What you have to do is, on your mobile device, open the latest version of the Google app or go to the Google Search widget, and tap the mic icon and say “what’s this song?” or click the “Search a song” button. Then start humming for 10-15 seconds. Using the Google Assistant, it’s just as simple. Say “Hey Google, what’s the song?” and then hum the tune.
After you’re finished humming, Google’s machine learning algorithm will identify the potential song matches. Google touts you do not worry about the perfect pitch to use this feature. Once you hum to it, the machine learning will come up with the “most likely options” matching to your hum. You can select the best match and give your ear and head some relaxation. Each option also shows the name of the artist, the name of the song, and the percentage of the match. The higher the percentage, the higher is the chance that the song you hummed to it is right.
Google also briefly explains how its machine learning model identifies the melodies or even whistles of the songs. Google says the system takes the song’s melody like its fingerprint. The machine learning model has been trained in such a way that can match your hum, whistle, or sing to the tight “fingerprint.”
When you hum a melody into Search, our machine learning models transform the audio into a number-based sequence representing the song’s melody. Our models are trained to identify songs based on a variety of sources, including humans singing, whistling, or humming, as well as studio recordings. The algorithms also take away all the other details, like accompanying instruments and the voice's timbre and tone. What we’re left with is the song’s number-based sequence or the fingerprint.
Machine learning then compares these sequences to thousands of songs from around the world and identifies potential matches in real-time.
The feature is currently available in English on iOS and in more than 20 languages on Android. Google promises that this feature will come in more languages in the future.