Sound sensor from Virginia Tech uses machine learning to more accurately interpret incoming sound waves.
Benefits
- Reliable
- Autonomous
- Accurate
Applications
- Medical implants
- Surveillance
- Security
UN Sustainable Development Goals Addressed
-
Goal 9: Industry Innovation & Infrastructure
The Challenge
Traditional sound sensors are often designed simplistically, to streamline the interpretation of the sensory outputs. Unfortunately, many sensors are unable to accurately identify the source of a noise because they are not good at eliminating extraneous information.
Innovation Details
The sound sensor consists of a synthetic bat ear and a microphone. The synthetic bat ear flutters as it hears a sound, funneling the sound waves toward the microphone. This microphone is connected to a system with a deep neural network, capable of learning sounds. The sensor is mounted on a rotating rig with a laser pointer, so when it “hears” a sound, it immediately rotates and points to it. The sensor can pinpoint the noise within half a degree, more accurately than humans, who are capable of pinpointing a sound within seven degrees.
Biomimicry Story
Bats use echolocation to pinpoint the location of prey and obstacles as they navigate through the air. To echolocate, bats send out sound signals from their mouth and nose. The sound signal travels through the air, bounces off the object, and returns to the bat, informing the bat of the location of the object. Essentially, bats use their ears to “see” their environment, much like how ships use sonar.