Localized Occupant Observing Camera

Project

Meetings and classes are commonly performed virtually.

• Cameras don’t dynamically follow the speaker.
• Microphone array and beamforming algorithms can be used to perform angle estimation.

The LOOC device follows the speaker.

• The device finds the direction of the speaker.
• The device moves the camera to point at the speaker.

System

Methods

• The Localization algorithm uses beamforming algorithms to estimate the angle of arrival of the sound.
• We take the angle of arrival from multiple pairs of microphones to find the azimuth and elevation of the angle of arrival.
• Given the estimated source, the motor controller then rotates the microphones toward the audio source (azimuth) and it rotates the mount up toward the audio source (elevation).
• At any point during operation the mode of operation can be adjusted to manual control, ignoring audio sources, the motors will turn as commanded from user entries.

Conclusion

• In simulation, the Localization algorithm tracks incoming single source signal with an accuracy of +/- 3 degrees.
• Hardware implementation in simple listening environments with a single source and a single path proved to be accurate.
• Topics Covered
  • Synthesize Complex Numbers in FPGA, Implementing Advanced Signal Processing in Hardware
  • Stepper and Servo Motor Control, Bluetooth Communication, Mobile App Development, 3D Modeling
• Future Work
  • Source separation for complex tracking environment
  • Dead zones for user customization
  • Echo cancelation for multipath