Studies in Signal Processing Approaches for Infrasound Wind Noise Reduction

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Wednesday, 7 January 2015: 5:15 PM
131AB (Phoenix Convention Center - West and North Buildings)
David L. Pepyne, University of Massachusetts, Amherst, MA; and A. Annan and D. J. McLaughlin

Infrasound is very low frequency sound (<20Hz). There are many natural and manmade sources of infrasound ranging from hurricanes, to tornadoes, to wind turbines, to sonic booms and explosive detonations. Since infrasound is generally recorded in an out-of-doors setting, one of the biggest challenges in infrasound signal detection is removing the portion of the recorded signal that is due to the wind. This infrasound wind noise, which is caused by turbulent eddies advecting through the boundary layer between the atmosphere and the ground, can more than double in spectral amplitude with each meter-per-second increase in wind speed to quickly overwhelm any infrasound signals that might be present in the recording. The usual way to mitigate the effects of wind noise is by attaching an infrasound sensor to a summing junction at the center of a spatial wind filter. A spatial wind filter, which consists of a circularly symmetric arrangement of pipes or hoses spread out over an area 15 or more meters in diameter around the infrasound sensor, provides signal-to-noise enhancement through constructive interference of the infrasound signal, which behaves as a train of coherent plane waves passing over the filter, and destructive interference of the random pressure contributions of the myriad turbulent eddies distributed over the area covered by the filter. Given that state-of-the-art spatial wind filters are only able to provide an order of magnitude reduction in wind noise, it can be necessary in high wind settings, such as the U.S. Midwest where infrasound has been tried for tornado early warning, to surround the spatial wind filters with an eddy fence to lift the boundary layer and break up the size of the eddies that reach the filter (since the frequency of the wind noise is inversely proportional to eddy size, breaking up the eddies attempts to move the wind noise towards frequencies above the infrasound band of interest). Seeking to avoid the severe deployment restrictions inherent in costly, large physical size, difficult to maintain eddy fence enclosed spatial wind filters, we have begun a study seeking to develop infrasound wind noise reduction methodologies based on applying various signal processing techniques to the recordings from easy to deploy, physically compact infrasound sensor setups. After an overview of the infrasound wind noise problem, we describe our infrasound sensor system equipment, the simplicity of its setup and operation, and the signal processing methods we are exploring for infrasound wind noise reduction. Results are then presented to show that, for certain classes of infrasound signals, such as the infrasonic tones emitted by large industrial wind turbines, our signal processing methods have the potential to be as effective as the spatial wind filter, and moreover, that our signal processing methods can be used along with spatial wind filtering to realize their combined benefits or to reanalyze pre-recorded spatially wind filtered data.