A group of MIT researchers has developed a new algorithm that improves upon the fast Fourier transform. In some situations, the algorithm is claimed to be 10 times faster than the existing function, which is good news for computing applications. The fast Fourier transform allows computers to decompose irregular signals into their component frequencies; it's used in everything from data compression to wireless signal processing, so a faster version has potentially wide-ranging applications.
The new algorithm is described in this paper, which is filled with equations that make my head spin this early on a Friday. As I understand it, the new approach is particularly applicable to "sparse" signals that have a small number of heavily weighted frequencies. Most of the "normal signals" in nature are sparse, according to one researcher, and the algorithm can better isolate their heavily weighted frequencies with a series of overlapping filters.
Audio and video compression routines seem likely to benefit from the new algorithm, and that's probably just the tip of the iceberg. We don't yet know how long it will take for actual implementations to be released into the wild, though.