LOUD: Large acOUstic Data Array Project

A 1020-Node microphone array and beamformer at the MIT Computer Science and Artificial Intelligence Laboratory.


Please see the data page for experimental data collected with the array.


Ubiquitous computing environments are characterized by an unbounded amount of noise and crosstalk. In these environments, traditional methods of sound capture are insufficient, and array microphones are needed in order to obtain a clean recording of desired speech. The LOUD (Large Acoustic Data Array) is a novel 1020-node microphone array utilizing the Raw tiled-processor architecture (TPA) for computation.

The LOUD modular microphone array currently consists of 1020 microphones. Our reasons for building such a large array are twofold. First, according to sensor array theory, the performance of a microphone array improves linearly as the size of the array grows. LOUD is currently the largest microphone array in the world [Guinness Book of World Records, 2007]. Second, ubiquitous computing applications often involve a large number of simultaneous feeds of streaming data (e.g., video, audio, haptics, etc). The I/O bandwidth and computational power necessary to process these streaming data has pushed the limits of traditional computer architectures and I/O schemes. To this end, our group has been designing a scalable polymorphic computer architecture called Raw, specifically designed to process large volumes of streaming data, such as that created by a microphone array. Our microphone array, which generates nearly 50 MB of data every second, is an appropriate target application for this architecture.


Prof. Anant Agarwal
Ken Steele
Eugene Weinstein



Pictures here


Microphone Array and Stereo Camera integration -- tracking two people as they move around. [Download] [Watch on YouTube]
Microphone Array - switching between several people all talking at the same time. [Download] [Watch on YouTube]

Computer Architecture Group

ecoder at-sign alum period mit period edu