Joint Magnetic Calibration and Localization Based on Expectation Maximization for Tongue Tracking

Area of Science:

• Biomedical Engineering
• Speech Science
• Robotics

Background:

• Tongue tracking offers insights into speech mechanisms, aiding speech therapy and language learning.
• Wireless localization using magnetic tracers provides a cost-effective method for capturing tongue kinematics.
• Accurate calibration of magnetic sensors is crucial for precise tongue tracking but is affected by environmental noise.

Purpose of the Study:

• To develop a joint calibration and localization (JCL) algorithm for wireless tongue tracking.
• To improve the accuracy of magnetic sensor calibration and tracer localization in speech research.

Main Methods:

• A Bayesian framework was used to model tracer movement and magnetic measurements.
• The Expectation-Maximization (EM) algorithm was employed for joint calibration and localization.
• The Unscented Rauch-Tung-Striebel smoother and curvilinear search algorithm were utilized for localization and calibration, respectively.

Main Results:

• The JCL algorithm achieved an average root mean square error of 0.45 mm for tracer position and orientation estimation.
• This accuracy is significantly better than methods using separate calibration and localization.
• Measurements were conducted on a tongue tracking system using a small magnetic tracer.

Conclusions:

• The JCL algorithm enhances the localization accuracy of wireless tongue tracking systems.
• This demonstrates a potentially high-precision method for tracking tongue movements.