The BBC love their esoteric, obscure electronic music. I think it all started with Sigur Rós in Planet Earth and proliferated from there. It’s even started creeping into BBC Radio 4 trailers now. The trailer for Will Self’s ‘A Point of View: In Defence of Obscure Words’ had just such a music bed. I decided that I would very much like to know what that music was.
The music isn’t the only mystery, by the bye. I’m not entirely sure what the title of the programme is either. The BBC reports it to be ‘In Defence of Obscure Words’, whereas Will Self, who I would assume to be the authority in this case, goes with ‘In Defense of Obscure Words’. I assume that ‘defence’ follows the same rules as other words similar to it, namely that the ‘-ce’ ending is used for the noun and the ‘-se’ ending is used for the verb. But I’m not entirely sure what ‘in defence’ (or ‘in defence’ for that matter) actually means. Is it some obscure conjugation of the verb, or is the ‘in’ a preposition to ‘defence’ as a noun?
But back to the identification of the background music.
I intended to use Shazam to try to and identify the track. I was able to find a white paper on Shazam (back when I was researching my dissertation) and, at least at the time of the paper, the fingerprinting is based on amplitude patterns rather than pitch content. As there was talking all over the track, I knew that wouldn’t work.
It didn’t stop me trying, however. I needn’t have bothered, it found nothing.
Radio is an interesting medium in that it’s designed to be mono with stereo enhancements. The story of stereo in the FM format is a whole fascinating story in itself. The net result is that things tend to be mono, with excursions into stereo for effect. The broadcast itself kept to this pattern: the voices were in mono (i.e. the signal was identical on both left and right channels) and the music was stereo (slightly different signals on each track).
Of course, if you subtract a value from itself, you end up with zero. If you subtract a number from a slightly different number, you don’t get zero. This is a recognised technique for removing vocals (which tend to be panned to the centre of the track) from background tracks. I decided to try this for myself.
First I found a broadcast on iPlayer which included the trailer. I used the brilliant Soundflower to direct output to an Audacity session. The result was this recording:
I then split the tracks into the two components and inverted one half. Inverting means multiplying the value by -1, so +10 becomes -10. This is interesting to listen to. If the signals were coherent, the voices would cancel out. If you’re listening through speakers try moving your head around in the field for a strange effect.
I then added the two tracks together, letting the identical components (speech), cancel each other out. The result is almost perfect removal of vocals (except where compression artefacts muddied things a bit and made the identical signals not-entirely-identical). It’s a bit of a hatchet job on the music (of course, some parts of it are in phase) but it’s still recognisable, and certainly still has the same amplitude profile.
Quite interesting! The music is still audible.
So, track identified? Problem solved? Well no, not quite. Shazam still has no idea.
Perhaps you do?