Google last year introduced a new ‘Now Playing’ feature with Pixel 2. This uses deep neural networks to bring low-power, always-on music recognition to mobile devices. Google recently introduced a new version of the Sound Search that is powered by the same technology as Now Playing.
Google also announced that you can use it through the Google Search app or the Google Assistant on any Android phone. Thanks to the latest version of Sound Search, Google says that you’ll get faster, more accurate results than ever before. Google developed an entirely new system using convolutional neural networks to turn a few seconds of audio into a unique “fingerprint without any noticeable battery impact.
The server-side Sound Search system is very different, having to match against nearly 1000x as many songs as Now Playing and this makes Sound Search both faster and more accurate with a substantially larger musical library presented several unique challenges.
Now Playing generates the musical “fingerprint” by projecting the musical features of an eight-second portion of audio into a sequence of low-dimensional embedding spaces consisting of seven two-second clips at 1-second intervals. In the initial phase, the algorithm finds good candidates by leveraging more neural networks and in the second matching phase, a detailed analysis of each candidate is performed to find the correct one.
Google for Sound Search, increased the density from 1s to every .5s to improve the match. Google also worked to lower the matching threshold of popular songs. Google concluding has said that it is working to improve matching in very quiet and noisy environments.