AudioMelody - make better music

How do music recognition apps work?

If you’ve seen the hit TV show Silicon Valley, you probably remember that Pied Piper, the game-changing compression algorithm, started out as a music-identification app that determined whether or not a piece of music was plagiarized.

Music recognition apps have been around since 1999, when Shazam began identifying songs. Since then it has been used to identify over fifty billion times. While they have become widespread, not many know how these apps actually work. What makes a music recognition app a good one? And which are the best?

How music recognition apps work

For a user, it’s fairly simple. First open the app, press record, and allow your phone to listen to the song. Just a few seconds goes by and the app will tell you what the song is, even if there is background noise or distortion. It almost seems like magic, but what it is really an effective algorithm that makes it possible.

Music identification apps work in the same way. With a large database of information about songs, the apps provide an interface to find exactly what song it is when you are curious. According to the website MoneyPug, which is used to compare mobile phones, you actually don’t even need a smart phone.

Basically it works like this. The database stores song data about the song’s unique sounds and patterns. Then the app listens to the music, creating an audio fingerprint based on a few seconds of the track. This data is the checked against the database, which matches it to your recording.

This begins with a spectrogram, essentially a graph that has time on the x-axis and frequencies on the y-axis, and amplitude is represented by color. Any sequence of sound can be converted into a spectrogram and there can be an assigned set of coordinates attached to it at any time. Notes turn into numbers. Looking through a database of millions of songs using a detailed spectrogram, there too many data points to look through at any speed.

Of course this was a breakthrough in music recognition. The realization that you can identify sounds with data changed how people view digitally identifying sound. Getting rid of a song’s lower energy parts decreases the spectrogram and makes the app less susceptible to identifying dull sounds and background noise.

Every second of a song is stripped to a few data points. The next step is to hash out this sequence of peaks. This means taking a set of inputs and running them through an algorithm that assigns an integer output, which is generated by taking two of the highest peaks and measuring the time between them, which puts together the two frequencies.

A string of numbers is the result, which are easily storable and searchable. A computer reads this has and recognizes it based on time and distance. When all of the songs data has been identified and hashed, the song now has a unique 32-bit number that becomes its ID in the database.

Which apps are the best?

Shazam is by far the most known app related to music recognition, but is it the best out there? Shazam has a simple interface but it has a “Discover” panel shows music news that includes new releases. The app enables you to hear previews or add songs to your favorites. It also allows you to access tons of information about the song. You can even listen to a sample, share it with your friends, and find the music online. What makes Shazam attractive is that it was designed for music lovers.

SoundHound is another one of these apps. It is similar to Shazam, but it has features like voice control and a Charts tab that helps you find features on its music player. Play samples, tag songs, read lyrics, and find the album the song is on. You can also sing to the app to identify a song. Finally, Musicxmatch is another competitor to Shazam. It is more a lyrics app. It helps you identify music by the lyrics instead of the song data. For music recognition apps, it’s all about the algorithm. Shazam has a lock on the industry, but as soon as a company comes up with a better algorithm, this will change. The other music recognition apps come at it from another angle, making them worthy competitors.


Tagged as:

One Response to “How do music recognition apps work?”


  1. How Do music Recognition Apps Work? #MusicMonday « Adafruit Industries – Makers, hackers, artists, designers and engineers!

Leave a Reply