Audio Systems Laboratory
Kevin D. Donohue,
The following links are to audio, image, and video files related to experiments and demonstrations with microphone array systems. The lab equipment consists of a cage (13’x13’x7’) on which speakers and microphones can be mounted for collecting or rendering acoustic energy over an array system. The cage was built with components from 80/20 Inc. The Industrial Erector Set. In addition, acoustic treatments can be mounted on the wall of the cage to realize various noise and reflectively properties of a variety of rooms. The data collection and processing is driven by 2 dual AMD processor computers running Linux. The data recording computer is connected to 2 Delta 1010 cards by M-Audio with 8 analog input channels each sampled up to 96kHz (24 bit) and 8 analog outputs each.
Pictures of Audio Cage:
Pictures of Electronic Interface (for 8 speakers and 8 microphones):
Acoustic image (steered response power (SRP)) from a 20 ms audio frame:
Sound Source Location movie of a walking guitar:
This is a quick-time format movie of SRP images of a guitar playing in front of an 8 element linear microphone array. The microphones are 40 cm apart and are located at the bottom of the movie frame. The field of view (FOV) is 3 meters wide and 2 meters deep. At 2 points the guitar steps out of the FOV; however because the coherence length of the tonal sounds is so long and the wavelength are as large as 2 to 3 meters, the sound will bleed into the FOV. In addition, with a 40 cm spacing, aliasing (for frequencies higher than about 430 Hz) and strong sidelobes can be observed in this image. No processing is done to reduce these artifacts: (Walking Guitar)
Video and Acoustic SRP movie together:
In these examples a perimeter array is used, consisting of 8 microphones, 2 on each wall arranged on vertices of an equilateral octagon with a spacing of 1.35 Meters at a height about equal to mouths of the speakers in the cage. A fisheye lens was mounted to the ceiling about 7.5’ above the floor and an angle of 185 degrees. So the audio SRP data was collected with the video so that the activity in the cage could be viewed while watching the SRP movie. The fisheye lens distorts the image nonlinearly over the acoustic imaging plane, so a one-to-one correspondence between the video and SRP image is not observed (except when the speaker was directly under the lens). For all movies the FOV was 3.225m by 3.225m.
This movie is a single speaker talking and moving around, the video image is on top and the SRP image on the bottom: Single Speaker Split
This movie is the same as above except images are overlaid: Single Speaker Overlay
This movie has two speakers talking simultaneously and moving around, the video image is on top and the SRP image on the bottom: Two Speakers Split
This movie is the same as above except images are overlaid: Two Speakers Overlay