post #1 of 1
Thread Starter 

Hi all,


I stumbled across this recent paper in the IEEE Transactions on audio, speech, and language processing journal and thought I'd share it because it is a peer-reviewed scientific publication on audio processing techniques (this is the science section of an audio forum, isn't it?). 


the article is:

Defraene, B.; Mansour, N.; De Hertogh, S.; van Waterschoot, T.; Diehl, M.; Moonen, M., "Declipping of Audio Signals Using Perceptual Compressed Sensing," Audio, Speech, and Language Processing, IEEE Transactions on , vol.21, no.12, pp.2627,2637, Dec. 2013
doi: 10.1109/TASL.2013.2281570




here's the link I found the PDF under:


(NOTE: You may need an active subscription to the journal to view the paper. I accessed the paper through my university's library, so you likely have access if you are on a university campus or at a business which subscribes to scientific literature)


Primer: Declipping is an audio processing technique by which a digital audio signal that suffers from clipping is "restored" to remove the artifacts created by clipping distortion. Obviously, hard clipping results in the loss of information; hence, no new information can be created by a declipping algorithm. However, declipping can be used to suppress the highly audible artifacts of hard clipping, which may lead to a subjective improvement of the signal.


I thought I would share this paper because it contains some signal processing theory, the design of a declipping filter, experimental tests of the technique's objective performance, and finally, experimental tests of the technique's subjective performance. The whole thing is pretty neat, because it explains the underlying principles of the problem (if you care to understand how things work) and then provides some subjective listening results (if you love statistics). There's a little something for everybody in here!