How Shazam Works

Under the Hood: The Audio Fingerprinting Technology Behind Shazam

Jan 13, 2024

“Time, as it grows old, teaches all things.” -Aeschylus

For anyone who doesn’t know. Shazam is an application that identifies the music used in movies, advertising, and television shows based on a short sample played. In this essay, I’m going to talk about how Shzam makes use of Audio Fingerprinting technology to identify music based on samples played.

How Does Shazam Work?

Shazam identifies songs through something called an audio/acoustic fingerprint and a Spectrogram. Now to explain these words.

What is an audio/acoustic fingerprint?

An audio/acoustic fingerprint is a condensed digital summary that is generated by an audio signal. An audio signal is a representation of sound typically using either a changing level of electricity voltage for analogue signals or a series of binary numbers for digital signals.

In simple words, an audio fingerprint is a unique digital summary of an audio signal. The audio signal is a digital representation of sound using binary numbers.

In the case of Shazam, these audio signals are a series of binary numbers used to represent digital signals, these binary numbers can be used to identify an audio sample or quickly locate similar items in an audio database (In case you aren't aware, an audio database is a database for audio).

What is a Spectogram?

A spectrogram is a visual representation of the spectrum of frequencies in a sound or audio signal. It is created by splitting the audio into small segments over time and running a Fourier transform on each segment. This separates the complex sound wave into its component frequencies and their amplitudes.

The spectrogram plots these frequency components against time on a graph. The x-axis represents time, the y-axis represents frequency, and colour or intensity represents amplitude/intensity. Lower frequencies are plotted at the bottom and higher frequencies towards the top.

So in essence, a spectrogram provides a way to visualize the change in frequency and amplitude of an audio signal over time. The unique pattern of the spectrogram acts like a fingerprint, enabling the comparison of different sounds and audio samples.

In simple words, a Spectogram is a coloured graph of sound which helps to discover the different structures of audio. To efficiently search for a sound you need to efficiently describe it and the way to do this is by using a spectrogram.

How does all this work in Shazam?

We've successfully explained the technologies used in Shazam as single concepts, now let's understand how they work together cumulatively to make up Shazam.

When you ask Shazam to tell you information about a song like its name, author, etc, you give it an audio stream of the song in question via a microphone or some other audio input device. It represents the audio stream as a spectrogram, the shazam algorithm then picks out the peak point in the audio stream via the spectrogram graph representation -Peak points are points of less background noise.

The Shazam algorithm then creates an audio fingerprint from the peak point and then indexes through the audio database for a song with a similar audio fingerprint and when it finds a successful match it then returns its results to the user.

How Shazam Updates its Audio Database

Going through this essay you might have been able to infer that a core piece of technology behind the success of Shazam is its extensive audio database, to put it simply without an up-to-date audio database Shazam won't efficiently meet the demands of its users. So how does Shazam keep its audio database updated?

They do this through Industry partnerships with companies who document music. Shazam gets these companies to document music for them and then uses the data it gets from these companies to improve its audio database.