Imagine being able to stand in the middle of a loud party or an airport with hundreds of people, and selectively single out voices that you need to analyze for security reasons. At the other extreme, imagine you are in a factory with many noisy machines running, and you can actually single out the noise from a specific device you are interested in.
All this can be achieved using a device invented by Siemens and University of Maryland researchers. This optimal sensor is capable of all this and more.
The invention that researchers decided to call "the optimal sensor" mainly cuts down on the number of microphones arranged in an array or network for the purpose of monitoring sounds in order to generate data and information about real time audio.
Radu Balan, professor at the Department of Mathematics and Center for Scientific Computation and Mathematical Modeling, said that the device would be useful to large industrial manufacturers or equipment manufacturers like Siemens as well as security companies. Justinian Rosca, senior key expert of Siemens Corporate Technology in Princeton, NJ, highlighted further that the device can be laid out in a variety of geometrical configurations and shapes to fit the application domain, unlike any other microphone array.
Balan, Rosca, and Mark Lai, a UMD Ph.D. graduate who is currently a postdoctoral fellow in The Institute for Computational Engineering and Sciences (ICES) at the University of Texas, Austin, have developed a method to determine the optimal way to arrange and place a microphone sensor network. This helps reduce both the number of microphones needed and the bandwidth requirements.
“We envision two typical setups,” Balan said. ”[First], the industrial setup, where the system monitors an industrial noisy area, for example, an electric power generation station where many engines and turbines run and produce sounds.”
This would involve acoustic case sensors, microphones, as well as accelerometers for vibrations, thermometers for temperature, current and voltage, and maybe even infrared cameras.
The other scenario is a security setup, ”where an array of microphones ‘listens’ to an area, e.g., airport waiting areas, and searches for specific ‘signatures’ [such as keywords, or voice prints],” he said.
The technology could monitor factory equipment, build security applications, as well as enhance oil and gas discovery, said Mark Lai, who was Balan’s graduate assistant when the research was going on.
So how is this sensor system better than existing ones?
“Our algorithm harnesses the power of multi-sensor arrays. We not only tell you where to place the sensors or microphones, we also tell you how to process the combination of received signals to extract the desired sound,” Lai said.
The method that Balan, Rosca, and Lai developed uses the convex optimization technique, which can help determine the ideal subset of sensors needed so that the interference gain, from the huge number of audio sources being monitored by an array, is kept at a minimum.
Getting this device to work the way they wanted wasn’t easy.
“Our first challenge was to define a realistic use case,” Balan said.
Siemens, who sponsored the initial research, came around to help them define scenarios of use that would interest the company. Another challenge researchers encountered was to get access to real world data so that they could test the technology.
“We made only partial progress on this issue; we decided in the end to use simulated data and calibration information,” Balan said. Another interesting conundrum was the question of how to control the algorithm complexity and make it work on available computing platforms. “Mark Lai is the ‘wizard’ that made this possible,” Balan said.
The Office of Technology Commercialization (OTC) helped Balan and his team navigate smoothly through the patenting and licensing process.
“OTC helped us with defining a business case for a related grant activity [NSF I-Corps], and in patenting the solution. Alla McCoy was instrumental in both these activities. Additionally she helped us navigate the IP issues with our industrial sponsor [Siemens],” he said.
This technology is available for licensing. For more information, contact email@example.com.
September 24, 2015
UMD Researchers Build a Smaller Device to Pick up Bigger Noises
Did You Know
UMD's Neutral Buoyancy Research Facility, which simulates weightlessness, is one of only two such facilities in the U.S.