R&D News & Events

Breaking the Search Engine Sound Barrier

Newly released software is taking on the last challenge in audio searching. The tool, called MediaMined, is an artificial intelligence system that can make sense of what it hears, whether the sounds are speech, music, or even a sound effect like an explosion or creaking door.

These days, it’s easy to take for granted the power of online text searches. With Google and Bing able to find just about anything written online, we may not appreciate how tricky it is for computers to search through pictures, video, and sound. Unlike people, computers cannot easily recognise objects in pictures or identify sounds heard in an audio file. One solution to this problem has been to label files with keywords, but the ideal solution would be to create software that understands content without needing such help.

Created by San Francisco-based Imagine Research and supported by funds from the US National Science Foundation, MediaMined is not the first tool designed to ‘understand’ audio. Voice recognition technology enables software to digest speech andvtools like Shazam and Soundhound’s Midomi and Hound have also come on the scene in recent years and can recognise music. MediaMined, however, extends audio recognition capabilities to all sounds, earning it applicability beyond speech and music.

MediaMined sets itself apart from other sound-searching tools by applying a machine-learning approach. This flexible strategy lets users find sets of similar sounds based on features beyond what might be in the keywords associated with a file. As Imagine Research’s founder and CEO Jay LeBoeuf explained in a recent press release, MediaMined “allows computers to index, understand and search sound- as a result, we have made millions of media files searchable.” With its general applicability, MediaMined could help movie soundtrack makers work more efficiently or, its creators speculate, even help doctors to assess a patient’s cough or wheeze.

Posted on Wednesday, 23rd November, 2011