Microsoft applied for a patent on an "automatic censoring filter" in 2004, according to Ars Technica. In the technology, speech-recognition software scans what's being said in real-time and distorts the audio stream if it becomes probable that obscene language is about to be used.
The gaming corner of the internet is quick to speculate that the technology will come to the profanity-laden Xbox Live, but an application in greater demand of the trick is television. Once brought to market, the technology is expected to be sold to television networks, who could rest easy knowing they're automatically avoiding fines by the FCC for airing indecent content.
Ars speculates that what satiates an easily offended American television audience could also be used by governments easily offended by politically-motivated speech.
In short, stopping 12-year-olds from screaming racial epithets during Halo 3 matches is the tip of the iceberg--and probably not the main reason--for Microsoft's patent. The science behind the tech, reproduced from the patent application, follows:
The automatic censoring filter employs a lattice comprising either phonemes and/or words derived from phonemes for comparison against corresponding phonemes or words included in undesired speech data. If the probability that a phoneme or word in the input audio data stream matches a corresponding phoneme or word in the undesired speech data is greater than a probability threshold, the input audio data stream is altered so that the undesired word or a phrase comprising a plurality of such words is unintelligible or inaudible.