Detection system uses acoustic sensors

Andrew Wilson, Editor, [email protected]

Data fusion from acoustic sensors, digital maps, global positioning systems (GPSs), and other types of sensors is of special interest to designers of military systems that must rapidly monitor or track specific threats. At the recent SPIE Defense & Security Symposium (April 2007; Orlando, FL, USA), Information System Technologies (ISTI; Fort Collins, CO, USA; www.infsyst.com) demonstrated an antisniper system. Rather than using an infrared camera to locate the position of a potential sniper, the ISTI system has low-cost distributed acoustic sensors that utilize ZigBee-based wireless networking.

Combining these fused data with sensor-location information on a host computer allows the operator to visually locate any potential sniper threat within seconds of its occurrence. According to Mo Azimi, president and CEO of ISTI, this distributed collaborative processing provides an effective means of capturing muzzle blasts and/or ballistic shockwaves while overcoming line-of-sight and multipath problems common in urban environments. “Moreover,” he says, “such systems offer stealthy operational capability and the ability to handle multiple sniper shots without confusion.”

Using 15 randomly distributed nodes (inset), a host PC displays the results of four different sniper positions overlayed onto a digital map with a localization accuracy of better than 1.2 m.

ISTI teamed with designers at Vanderbilt University (Nashville, TN, USA; www.vanderbilt.edu) to develop a combined vehicle- tracking and transient-event-detection sensor board that features five analog input channels. “Four of these channels use 12-bit analog-to-digital converters (ADCs) and sampling resolutions that provide the 1-µs temporal resolution required for gunshot detection and localization,” says Kumar Srinivasan, a member of the Technical Research Staff at ISTI.

Using an on-board Spartan 3L FPGA from Xilinx (San Jose, CA, USA; www.xilinix.com), these channels are operated in a synchronous fashion. The fifth channel, however, uses a 16-bit ADC and operates at 100 kS/s, used primarily for vehicle tracking. The board also uses a Chipcon CC1000 radio that provides the added capability for sensor self-localization with phase interferometry. The analog channels can also be used for other sensor modalities such as seismic and low-resolution IR.

The sensor board can be interfaced to a TelosB mote from Crossbow Technology (San Jose, CA, USA; www.xbow.com) over an I2C interface. This mote provides the system with an IEEE 802.15.4 radio interface from which captured and processed data can be transmitted to a PC-based wireless base station.

“To localize a specific transient event,” says Srinivasan, “the sensors need to determine the onset of an event, the presence of a militarily significant transient signature, and its duration.” The presence of a transient event in signals received by all the sensors is determined using a power-based detection method. At each sensor, a number of samples are taken and the energy of the recorded signal computed in the FPGA board. To detect the presence of a transient signal, the FPGA checks whether two consecutive snippets have energy greater than a predetermined value. The end of the transient signal is computed by checking if three consecutive samples have an energy of less than a second predetermined value.

“In this way,” says Srinivasan, “the method detects the existence of an acoustic transient signal and the onset and duration of the signal.” To form a complete sensor system, the sensor board is interfaced to the TelosB mote, five WM-64 omnidirectional microphones from Panasonic (Secaucus, NJ, USA; www.panasonic.com), an external antenna, and a battery pack (see figure). “The ISTI sensor board provides substantial sensor-level computational power that cannot be offered by the mote alone,” Srinivasan says.

The entire distributed network, consisting of several sensor nodes, is controlled globally using a MATLAB-based GUI from The Math Works (Natick, MA, USA; www.mathworks.com) that runs on a PC at the base station. After the synchronized time series of transient events arrive at the base station, a localization algorithm converts the corresponding time difference of arrivals to the event location using the GPS location of the sensor. GPS-encoded map data on the host computer are then used to display the location of transient events relating to sniper activity.