SmarTeeth: Augmenting Manual Toothbrushing with In-ear Microphones

Research Background and Problem

• Identified Problems or Challenges:
  Improper toothbrushing practices remain a widespread oral health issue globally, contributing to conditions such as tooth decay and gum disease. Despite high-end electric toothbrushes with brushing tracking features, manual toothbrushes remain more popular due to simplicity and cost, leaving a majority of users without proper brushing guidance.

• Significance:
  Oral hygiene directly affects public health, with diseases like periodontal disease and untreated tooth decay impacting millions worldwide. Enhancing manual toothbrush usability with smart tracking capabilities can bridge the gap for users lacking access to high-end devices.

• Research Motivation and Related Work:
  Existing solutions such as camera-based systems and IMU devices on toothbrushes have privacy concerns, sensitivity to user movements, or lack of fine-grained monitoring. Prior audio-based methods, while promising, struggle with ambient noise sensitivity. The emergence of in-ear microphones on Active Noise Cancellation (ANC) earbuds presents an opportunity to leverage bone-conducted brushing sounds for monitoring, addressing these challenges effectively.

Solution

• Proposed Solution:
  The authors present SmarTeeth, a system utilizing in-ear microphones on earphones to capture bone-conducted brushing sounds for fine-grained tooth surface tracking. This system extends functionalities typically seen in high-end electric toothbrushes to manual toothbrushes.

• Innovations:

  • Use of bone-conducted brushing sounds captured via in-ear microphones.
  • Development of cross-channel features (coherence and phase) that inherently characterize distinct toothbrushing surfaces.
  • Integration with deep learning for accurate tracking of brushing regions and surfaces.
  • Real-time auditory feedback via earbuds to enhance brushing practices.

• Implementation Steps:

  1. Sound Preprocessing: Filtering noise and detecting brushing moments using short-time energy analysis.
  2. Feature Extraction: Deriving both traditional audio (e.g., MFCC) and cross-channel features representing the differences in sound propagation pathways between ear canals.
  3. Classification: Using a deep learning model to fuse audio and channel-related features for surface recognition.
  4. Postprocessing: Applying temporal smoothing to refine predictions and reduce misclassifications.
  5. User Registration: Fine-tuning the system for new users with minimal calibration data (e.g., 3 sessions of 2-minute brushing cycles).
  6. Feedback Mechanisms: Providing underbrushing and overbrushing alerts, brushing score computation, and area-specific brushing duration visualization.

Research Outcomes

• Specific Results:

  • Achieved an average accuracy of 92.7% for 6-region identification and 75.6% for 16-surface detection after one registration session.
  • Accuracy improved to 98.8% (6-region) and 90.3% (16-surface) with three registration sessions.
  • Demonstrated robustness across different brushing orders, toothbrush types, and environmental noise levels.
  • Successfully extended methodology to low-end electric toothbrushes, reaching 92.4% accuracy.

• Advantages:

  • SmarTeeth compares favorably against baseline systems like BrushBuds (IMU-based approaches failed to track finer surfaces effectively).
  • Outperformed similar electric toothbrush tracking systems like ToothFairy by 6.4% for quadrant tracking, demonstrating methodological robustness.

• Experiments and Evaluation Results:

  • Performed detailed user studies, showcasing improvements in brushing scores when real-time auditory feedback was enabled.
  • Plaque tests validated system's effectiveness in monitoring brushing coverage.
  • In-the-wild experiments achieved highly accurate detection results, proving SmarTeeth's viability in realistic settings.

• Limitations and Future Directions:

  • Hardware Improvement: Current prototype is bulky and less comfortable. Collaboration with earphone manufacturers for miniaturized hardware is suggested.
  • Toothbrush Degradation: Real-world performance may fluctuate as bristles wear over time, necessitating periodic recalibration.
  • User Applicability: Future adaptations are needed for users with unique dental structures, such as braces or missing teeth.
  • Potential Applications: Expanding functionalities to detect brushing pressure and early signs of dental diseases like cavities and root canal infections.

Conclusion

SmarTeeth is a groundbreaking solution that leverages earphone-based acoustic sensing to enhance manual toothbrush usability. By monitoring brushing regions and surfaces in fine detail, the system empowers users of manual and low-end electric toothbrushes to improve their oral hygiene effectively. With promising experimental results and positive user feedback, SmarTeeth demonstrates significant potential to improve global dental health, while future development can extend its capabilities and address current limitations.