Can Your Car Hear Me Now?

Kyle Logue takes the audio sensors for a spin while Jason Zheng records data. (Photo: Heather Golden)

The development of autonomous cars is rolling along, and one of the issues to tackle is how these cars detect and identify obstacles in their surroundings. Aerospace’s Kyle Logue is proposing audio sensors as a potential low-cost method to help address this problem.

“Audio mics are an untapped sensor in the autonomous vehicle market,” he said. “A concise processing solution would allow high-accuracy positioning of unknown signals for object avoidance, collision warning, blind spot detection, and a general purpose sensor for autonomous driving.”

Other methods of detection are available, such as radar and lidar, but they can be expensive or they have limitations. Audio sensors are significantly cheaper, and have the potential to fill some of the gaps left by other sensing methods.

Logue has performed several field tests to collect data for his idea. He started by putting static microphones at set intervals to collect sound and calibrate his localization algorithms.

Then he taped eight microphones to his car and drove it around in circles in Aerospace’s parking lot while emitting sound from a speaker at a known location (see video of test below).

The equipment Logue used is commercially available and inexpensive. The trick is in processing the data, and Logue is developing algorithms that will turn that data into usable information.

“The primary obstacle to such a system would be the signal filtering stage given that roads are very noisy environments and that sensors themselves would be exposed to significant interferers like engine, exhaust, and road noise,” he said.

Regardless, Logue anticipates that audio sensors could be used to detect other cars, pavement types, tunnels, and more. The sensors can detect actual sound emitted from a source, like a car horn, or the way the sound bounces off a surface, such as a wall.

He also thinks the sensors could pick up on internal noises from the vehicle itself. “This would enable early fault detection for things like timing belts, wheel wear, axle bearings, and other identifiable signatures,” he said.

Another potential use is to record the audio data, which could then be used to analyze an accident for insurance purposes.

The possibilities are definitely intriguing, and Logue is working to make them a reality.

“The addition of multiple low cost audio sensors for localization of other vehicles, situational awareness, fault diagnosis of common problems, and recordings for crash reconstruction would be a cheap and valuable addition to any future vehicle,” he said.

Before putting the sensors on an actual car, Logue and Zheng collected some preliminary data with a stationary setup. (Photo: Kyle Logue)

Before putting the sensors on an actual car, Logue and Zheng collected some preliminary data with a stationary setup. (Photo: Kyle Logue)

Check out the video below

—Laura Johnson