Ears in the Sky:
SwRI accomplishes airborne acoustic surveillance via unmanned aerial vehicles
By Steve Cerwin
Battlefield commanders, intelligence analysts and law enforcement officers constantly seek new and better ways to gather overhead imagery, whether from satellites, reconnaissance airplanes or helicopters. However, significant intelligence also can be gathered not just by what a sensor can see, but also by what it can hear.
The military, security and law enforcement communities recently have become interested in collecting acoustic data from airborne platforms, particularly from unmanned aerial vehicles (UAVs). An airborne platform offers a significant advantage over ground-based mobility sensors: It can be deployed remotely to quickly and easily collect data over large areas. This allows the military to acquire remote battlefield intelligence from a safe distance, and it offers a unique data collection tool for security specialists. Typical applications could include locating gunfire and vehicles and identifying specific targets through their acoustic signatures.
Current methods for airborne acoustic sensing usually employ just two or four microphones to determine the sensor's bearing to the sound-emitting target. While these can provide line of bearing data to very loud sounds like gunfire, the resulting poor signal-to-wind-noise ratio limits their ability to record good position or signature data on weaker sounds.
In an internally funded research program, Southwest Research Institute (SwRI) designed, built and flew an airborne microphone array that applied more sophisticated technology to obtain improvements in several areas. SwRI engineers increased the number of microphones to reduce wind noise and improve bearing accuracy through averaging. They also extended the microphone arrays in two axes to provide both vertical and horizontal vectors to the sound source. This two-axis data, when combined with flight position information, can locate an emitter on a topographical map.
A two-axis, multi-element microphone array was installed on an electric-powered, radio-controlled (RC) glider. The SwRI team built two additional arrays, with sizes two and four times larger than the glider array. The largest array was installed on a commercially available parafoil aircraft modified as a UAV. The parafoil UAV's large aperture was intended to provide narrow beam widths at the low frequencies anticipated for vehicle sounds. Microphones were clustered to help average the acoustic interference from wind noise. Special foam shields were installed over each microphone station to reduce wind noise.
The SwRI team compared the effects of different array sizes in ground tests, where they collected simultaneous acoustic data on high-horsepower military vehicles.
The glider was hand-launched, then flown for 10-15 minutes per sortie for its acoustic tests. After climbing to altitude with its electric motor, it performed its data collection runs with the engine off to reduce noise while gliding at minimum controllable airspeed to reduce wind noise. The clean airframe afforded excellent wind penetration and proved effective in handling wind gusts up to 20 miles per hour.
Data-gathering flights by the parafoil UAV were flown over target ground vehicles at altitudes ranging from 500 feet to 1,000 feet. Operating with its engine off during tests, the parafoil UAV's large arrays provided excellent position tracking and signature collection from the test vehicles below. Both the glider and the parafoil UAV succeeded in gathering useful acoustical data during their respective tests. Those flight tests, combined with data gathered during ground tests, allowed the SwRI team to draw some significant conclusions.
Wind and engine noise are the two dominant obstacles to acoustic collection from a UAV in flight. Wind noise is generated by turbulent airflow over the microphone windscreens and over discontinuous features on the airframe. Engine noise during the data-gathering phase of each flight was handled by stopping the engine and then restarting it when the data gathering was completed.
The team found that wind noise could be mitigated by flying at minimum controllable airspeed; placing the microphones at acoustically quiet locations on the airframe; using a low-drag windscreen design; incorporating frequency shaping in the microphone preamplifiers to prevent low frequency overload; and using many microphones in the array in a configuration that allows directional listening and enhancing a desired signal while rejecting others.
The physics of an acoustic array were verified by the flight tests. The fineness of focus, expressed as narrowness of beam width of a steered array, is proportional to the physical size of the array in terms of wavelength. Larger arrays can produce narrower beam widths at lower frequencies. The three arrays progressively increasing in size by a factor of two were predicted to have the same beam width in progressively lower frequency octaves. This was verified in ground tests using the arrays simultaneously. The arrays showed identical beam widths, even though they came from arrays successively doubling in size and were processed in successively lower octaves.
A two-dimensional sound field plot can be obtained by using signal processing software to correlate the data from both the longitudinal and the transverse arrays. The DSP algorithms used to process the data provided two-axis sound field plots that updated in real time. In one data "snapshot," the sound field plot correctly placed the target vehicle slightly to the right and behind the UAV.
The processed array data proved capable of resolving the locations of more than one sound source simultaneously. Once, while vehicles on the test range were being swapped, a second vehicle started its engine and began to move while the first was still in motion. The airborne sensor array resolved the two vehicles in the sound field.
The large array flown on the parafoil UAV provided the expected improvements in directional sensitivity at low acoustic frequencies. However, because of wind noise, the dramatic improvement in signature quality that had been sought for low frequencies was not realized. Because the glider could fly slower than the parafoil, its data quality proved better than that from the parafoil. Even though the parafoil had four times the physical aperture, both vehicles had the same number of microphones and thus had the same amount of wind noise averaging. The glider's smaller array could not produce the narrow beam widths of the parafoil array, but the beam width it did produce was sufficient at the frequencies used in the tests.
Future directions for development
Coping with wind noise by increasing the number of data channels is difficult and expensive in terms of bandwidth and complexity. However, modern advances in signal acquisition hardware can allow use of even more microphone channels, improving performance accordingly. Acoustic and operational performance of both the parafoil and the glider might benefit from greater speed range, especially at the low end. However, the ideal vehicle may be a balloon, with zero air speed.
One possible data-collection scenario might be for a manned or unmanned aircraft to deploy several balloon-supported sensors over an area of interest, much as antisubmarine aircraft deploy sonobuoys in the ocean. The balloons could auto-inflate after release, then float in the prevailing air mass at zero relative velocity.
Using the technology developed for the internal research project, the SwRI team is prepared to adapt an existing UAV platform for acoustic collection service, or to develop aircraft or deployment scenarios specifically dedicated to this task to fulfill client needs.
Comments about this article? Contact Cerwin at (210) 522-2903 or firstname.lastname@example.org.
Published in the Winter 2004 issue of Technology Today®, published by Southwest Research Institute. For more information, contact Joe Fohn.