David Sun::Projects  
  Microphone Arrays for Hands-free Speech Interfaces next

Human-machine interaction of the future is clearly moving towards speech as a main interface. Many compelling situations exist where hands-free speech interaction with a system is preferred: think about the last time you had to make or take a call while driving on the highway. Many hands-free applications run in environments with high levels of acoustic reverberation and SNR loss. Current speech recognition systems are built primarily for desktop computing and thus lack robustness with respect to hands-free speech. Our work aims to tackle this problem.

The cost of supporting computing technology is already approaching the point where microphone arrays will be practical in many common environments such as conference rooms, lecture theaters, homes, and hospitals. Microphone arrays have the potential to significantly boost the signal quality required for building robust and usable speech-driven interfaces.

System Description

We have installed a large-aperture microphone array over an open-floor lab using commodity hardware:

  1. A total of 36 omni-directional microphones.
  2. Audio signals are sampled at 44.1KHz and streamed through six MOTU firewire audio interfaces in real-time to a server box.
  3. A time delay algorithm (PHAT) implemented in C/C++ runs on the server in real-time.
  4. A NLS localization algorithm implemented in Matlab/C++ runs asynchronously.
  5. visualization of sound sources implemented in Adobe/Flash will be accessible from browser interface.

Applications

The mic-array system will serve as the interface to the following applications:

  1. the ILLUMINAC speech interface to dynamically control the configuration of overhead lighting
  2. a visualization for speaker localization and tracking, and an interface that allows people working in private and shared space to mediate their voice levels

Conference papers

Coming soon!

Software

Coming soon!