He didn't get where he is today by stealing somebody else's catchphrase.

Lend Me Your Ears

If there's a sense that I wouldn't want to lose (in the unlikely event of being presented with a choice), it would be my hearing. For a music-lover such as myself, deafness would have a major impact upon my life. During my working day I listen to music for a good few hours, whether that be podcasts, Dandelion Radio, Spotify or my RavensPi. I’ve got the attention span of a goldfish (and marginally less memory-retention), so music keeps me focused. It keeps my concentration levels up whilst working, keeps me alert when driving, blocks out the constant babble of schoolchildren when commuting and drowns out my own heavy-breathing when I'm running. I even wear my headphones at the supermarket.

My favourite part of listening to music is the discovery of something new and unknown, which means that occasionally I'll use identification apps to find out about new tracks. Most people have only heard of Shazam, but there are others. Is there an app out there that does a better job? Does anyone else have a better catalogue or identify tunes with better accuracy? There’s only one way to test - and test with a diverse range of songs as only I know how.

So, with a playlist of twenty-five tracks I've tested three different apps - Shazam, Soundhound and MusicID. The apps were tested using the same sections of each song, played from the same source (my laptop's iTunes library) at the same volume and each app was given three attempts to find a match. The tracks spanned a variety of genres, years and popularity from the mainstream to the downright obscure. The twenty-five songs I've used are below. Of course taste is subjective, but hopefully you'll get the idea of what I was trying to achieve. The majority can be found on Spotify, but for those that aren't I've also put links to YouTube or BandCamp.

1) Prophecy Of Doom - Onward ever Backwards
2) The Orb - A Huge Ever Growing Pulsating Brain That Rules From The Centre Of The Ultraworld
3) Pink Floyd - Shine On You Crazy Diamond
4) Jon Hopkins - Abandon Window
5) Metrocross - Cowcube
6) Orbital - Doctor
7) dot - Neverland
8) Robson Ponte - Reinhard Voigt (Wassermann Mix)
9) Boards Of Canada - Left-Side Drive
10) Bomb The Bass - Mega Dis
11) Dan Le Sac vs Scroobius Pip - Letter From God To Man
12) Robert Mitchum - From A Logical Point Of View
13) Ulrich Schnauss - Goodbye
14) The Shoes - Crack My Bones
15) Steve Reich - Music For 18 Musicians - Part 4
16) Moderat - No. 22
17) Monty Python - Bruces’ Philosophers Song
18) Cuban Boys - The Nation Needs You (2014 Version)
19) The Future Sound Of London - Cascade
20) The Wedding Present - Corduroy
21) The Soundcarriers - Low Light
22) Blueneck - Pneumothorax
23) Sebastien Tellier - La Ritournelle
24) Bibio - Don’t Summarise My Summer Eyes
25) Happy Flowers - My Head’s On Fire

In some cases, I've deliberately picked tracks because they have samples (or sound like) other songs. For example, Number 11 has a whacking great sample of Radiohead's Planet Telex and Number 2 has a sample of Minnie Riipperton's Loving You. I've tried not to make it too easy. If I picked 25 mainstream tracks, all three would probably score close to the maximum.

So, how did they do?

With a score of 19 out of 25, Shazam is still the best music identification application out there.

That's not to say that Shazam doesn't have its faults. One thing I have found in general usage is that Shazam will identify tracks incorrectly sometimes, coming up with a completely different name and artist. One example that took place during this test was when it identified Blueneck as Enya. The only way to sanity-check an identification is by hunting for the track once it's identified and seeing if it was actually what it should've been - sometimes it isn't.

Probably the most disappointing app was MusicID, which is the only non-free app on the iTunes store out of the three. This all goes to show that you don't necessarily get what you pay for. I also noticed that in comparison to the other two its speed was generally slower.

However, music identification has got better. I remember in my pre-iPhone days that my Sony Ericsson W980i had such a feature, although it was a bit hit'n'miss. Things have certainly improved in the last seven years. Perhaps in another seven years, I'll be able to see apps that score significantly higher than these. Not all of us have mainstream music tastes and hopefully the providers of these services will find a way of getting around this problem when it comes to identifying songs - especially if they'd like us to buy them!
blog comments powered by Disqus