Solving the cocktail party problem.

The article discusses advances in automated speech recognition technology that would allow computers to separate two streams of words being spoken at once. According to the author, computerized speech separation could improve speech recognition systems, in addition to, advancing endeavors that require the separation of signals, such as understanding brain-scan images. Topics include an overview of the "cocktail party problem," which is the ability to pick a thread of speech from the conversation of two or more individuals and the challenges of creating a computer that can separate two streams of words being spoken at once.