IRENE: Key to unlocking mute recordings
Connecting state and local government leaders
Scientists from the Lawrence Berkeley National Lab and the Library of Congress are collaborating on using digital images to reproduce sound.
The idea is simple: Make a high-resolution digital image of a sound recording and develop software to analyze the image and reproduce the effects of a phonograph needle.
Scientists at the Lawrence Berkeley National Laboratory and the Library of Congress have brought the concept to a practical application over the past 10 years, developing a system to digitally recover and preserve rare and damaged recordings without the risk of doing additional damage.
Related coverage:
A race to restore the voices of the past
“The approach evolved naturally out of methods of optical metrology, pattern recognition, image processing and data analysis we use for physics research,” said Carl Haber, a high-energy physicist at Lawrence Berkeley.
The result is the Image, Reconstruct, Erase Noise, Etc., or IRENE, now being used by the Library of Congress. A version capable of two-dimensional imaging is used for disc recordings with lateral grooves, and three-dimensional imaging is used for cylinders with vertical groove modulation. The 2-D version also works for imaging and playing back optically recorded audio files, such as those on early experimental recordings and on some motion picture soundtracks.
The workstations are helping to preserve hundreds of recordings at the library’s audio-visual campus in Culpepper, Va., and are being used in research on preservation of historic one-of-a-kind recordings such as those held by the Smithsonian National Museum of American History.
The Berkeley Optical Sound Restoration Project is being conducted with funding from several government agencies, along with grants from private-sector associations supporting libraries and the humanities.
IRENE is evolving toward a fully automated system capable of handling imaging exposure, focus, start-stop, parameters and selection. Software then combs over the image to remove obvious faults such as dirt, scratches and wear. Algorithms track and determine the center point of the imaged grooves and “play back” the record with a virtual stylus, producing a digital .wav file without touching the original recording.
Although it is becoming more automated and user-friendly, its developers do not see a consumer market for IRENE. “This tool finds its home in laboratories, museums and libraries,” Haber said. In addition to the LOC installations, there is a system at the Lawrence Berkeley Lab, and another is expected to be installed at the University of Chicago’s South Asia Library in Chennai, India.
Although it is not likely to be available at Best Buy anytime soon, IRENE still has a lot of work left to do, Haber said. “There are a dozen major sound collections or more around the world. We have only scratched the surface in terms of the major historical collections.”