Just today UPS dropped off (well, I intercepted the driver in mid-flight, but that’s a different story) the Jawbone Bluetooth headset I was waiting for – and boy, does it work wonders! This has to be the best implemented piece of DSP technology I have seen. If you don’t believe in ‘official’ demos and blurb from jawbone.com (they are true), just check out this audio recording [WAV, 48kHz, 1.7MB]. The first part is recorder using my Mac with an unsquelched radio about 30cm from the headset, with the DSP turned off. I then turn it on, and the audio is recorded perfectly, with barely a hint of the background noise.
How do they do it? I’ve not opened this baby up, as it’s too good-looking (and expensive), but it seems that they combine two directional microphones with a vibration sensor (the small white dot you see on the picture) that picks up bone-transmitted voice from your jaw (thus the name). By matching the vibrations, which are not enough to actually record sound, with the incoming audio from the microphones, they can take away the extra noise very effectively. I should know, as in my previous job, one of our biggest problems was noise and echo cancellation (I was responsible for electronics R&D at SouthWing, designing and testing Bluetooth headsets and other accesories) – and we could never completely kill feedback echo, and noise – forget it. Our designs were in the top league as far as audio quality and noise went, but try what I just did today and the whole recording would have been like the first half.
Good job, Aliph!