The “cocktail party problem” has long been a challenge for technology to replicate the human ability to filter out background noise and focus on a single conversation. However, a breakthrough in AI technology by Wave Sciences has now made it possible to separate overlapping voices in audio recordings.
Keith McElveen, the founder and chief technology officer of Wave Sciences, spent years developing an AI algorithm that can analyze how sound bounces around a room and suppress any sound that couldn’t have come from where the target speaker is sitting. This technology has already been used in a US murder case, where it provided crucial evidence for convictions.
The potential applications of this technology are vast, ranging from forensic investigations to military use in analyzing sonar signals. Wave Sciences aims to introduce tailored versions of its product for use in various devices, including smart speakers, cars, and virtual reality interfaces.
The success of Wave Sciences’ algorithm has not only revolutionized audio forensics but also shed light on the similarities between the algorithm’s math and human hearing. This discovery suggests that the human brain may be using similar mathematical principles to solve the “cocktail party problem,” opening up new possibilities for understanding how we process sound.
