SEOUL — Korean researchers say they have discovered a better way to detect porn by filtering content through sound.
Rather than looking at video imagery, which has a high 90 percent accuracy rate, researchers from the Korea Advanced Institute of Sciences and Technology keyed in on audio from pornography, which is distinct and easy to detect.
Researchers Hoirin Kim and MungJon Kim created spectograms of audio clips using a radon transform to represent the voice, music and sound effects along a parallel path.
Results proved a clear distinction between porn audio, which is high-pitched, quick changing and repetitive, and that found in other forms of media, including music.
When tested across multiple soundtracks, the Kims achieved a 93 percent success rate and believe they have found a way to increase detection rates and lower false positives.
The Kims noted that by using sound, they avoid the problem of visual porn-identifiers (pornifiers) that can get tricked by any expanse of skin, like closeups of the face or other not-inappropriate body parts.