FaceOri: Tracking Head Position and Orientation Using Ultrasonic Ranging on Earphones

Document Title

FaceOri: Tracking Head Position and Orientation Using Ultrasonic Ranging on Earphones

Document Information

• Topic Area: User Interface Design and Interaction Technology
• Keywords: Face orientation tracking, ultrasonic ranging, earphones, head pose estimation, human-computer interaction, wireless device interaction, active noise-canceling headphones, attention detection
• Conference and Year: CHI 2022, Human-Computer Interaction Conference
• DOI: 10.1145/3491102.3517698

Research Background and Problem

• Identified Problems or Challenges:

  • User face orientation is often indicative of interaction targets, but existing tracking methods (e.g., camera-based head orientation estimation and eye tracking) face privacy concerns, visual field requirements, and are unsuitable for non-camera devices.
  • Current ultrasonic ranging methods experience performance degradation due to non-line-of-sight (NLOS) paths and Doppler effects caused by rapid head movements.

• Importance:

  • Tracking users' head position and orientation enables more natural interactions, such as convenient touchless device control, intelligent context-aware interface design, and activity tracking functionalities.

• Research Motivation and Related Work:

  • Enhancing the potential of earphones as spatial input devices by leveraging their built-in microphones for face orientation and distance measurement.
  • Attempting to develop solutions without requiring extensive multi-device setups, building upon background techniques like Doppler positioning and FMCW ultrasonic ranging.

Solution

• Proposed Solution:

  • FaceOri is an ultrasonic ranging method based on earphone microphones, designed to track users' head orientation (including pitch and yaw angles) and spatial distance from sound-emitting devices.
  • It utilizes speakers in existing devices to emit inaudible ultrasonic signals, received by earphone microphones for ranging.

• Innovations:

  • No additional hardware required, leveraging commercially available ANC earphones and speaker-equipped computing devices for head orientation estimation.
  • Optimized FMCW ranging technology to address issues caused by non-line-of-sight (NLOS) paths and Doppler effects.
  • Creatively employs non-invasive techniques to achieve calibration-free attention detection.

• Implementation Steps and Key Techniques:

  1. Audio Signal Processing: Using FMCW ranging technology, earphone microphones receive ultrasonic signals emitted by speakers and calculate time-of-arrival differences.
  2. Geometric Modeling: Utilizing the distance from the head center to the microphones, pitch and yaw angles are calculated using triangle-based mathematical formulas.
  3. Optimization Strategies: Triangularly modulated ultrasonic signals reduce Doppler effect impact, and CFAR adaptive algorithms predict frequency peaks.
  4. Binary Attention Detection: Without geometric calibration, a classifier is trained to extract features from audio signals to identify whether the user is facing the device.

Research Outcomes

• Specific Results:

  • Distance measurement accuracy with a median absolute error of 10.9 mm, tracking errors for yaw and pitch angles of 3.7° and 5.8°, respectively.
  • FaceOri achieved a binary attention classification accuracy of 93.5%.
  • Compared to the CAT baseline method, FaceOri significantly reduced frame drop rates and improved robustness against NLOS paths and rapid movements.

• Advantages Over Existing Solutions:

  • Avoids privacy concerns associated with camera-based technologies and is applicable to non-camera devices.
  • Reduces heading drift issues compared to IMU solutions while achieving superior overall accuracy.
  • Optimizes ranging performance under NLOS paths and rapid movements compared to traditional FMCW methods.

• Experimental or Evaluation Results:

  • Experimental data demonstrated that FaceOri maintains high robustness against noise and positional errors when users are close to the device or rotating rapidly.
  • Tests under varying relative heights and environmental noise conditions showed excellent performance even in high background noise environments.

• Limitations and Future Directions:

  • Limitations:
    • Calibration is required to establish reference points for continuous tracking.
    • Current implementation relies on external processing devices (e.g., PCs) for real-time signal processing.
  • Future Directions:
    • Explore more convenient calibration methods, such as integrating Bluetooth time synchronization protocols or utilizing cameras.
    • Enhance deployment generality by evaluating compatibility across different earphone devices.
    • Expand support for multi-device interaction, investigating more efficient frequency or signal encoding schemes.

Summary: FaceOri demonstrates the feasibility of achieving head orientation and distance tracking using commercial earphones and devices, with high robustness and potential applications in intelligent interaction, activity detection, and attention-driven interface design. Future work could focus on fully user-friendly calibration methods and multi-device support.