Estimating Three-Dimensional Body Orientation Based on an Improved Complementary Filter for Human Motion Tracking.

Area of Science:

• Robotics and Wearable Technology
• Sensor Fusion and Signal Processing
• Human-Computer Interaction

Background:

• Rigid body orientation using Inertial Measurement Units (IMUs) is crucial for robotics, navigation, rehabilitation, and human-computer interaction.
• Existing methods often face computational challenges for efficient, wearable motion tracking.
• Accurate fusion of sensor data, particularly quaternions from angular rate integration and sensor-based algorithms, is key.

Purpose of the Study:

• To propose a novel quaternion-based complementary filter algorithm for computationally efficient, wearable motion tracking.
• To dynamically fuse quaternion data derived from angular rate integration and a FQA algorithm.
• To enhance the accuracy and real-time capabilities of motion tracking systems.

Main Methods:

• A gradient descent method was employed to derive a function from sample points.
• This function dynamically estimates a fusion coefficient based on deviations in magnetic field and gravity vectors.
• A custom test machine was developed for performance evaluation of the proposed filter.

Main Results:

• Experimental validation confirmed the effectiveness of the designed quaternion-based complementary filter.
• The filter demonstrated accurate dynamic fusion of sensor data.
• The system showed significant potential for real-time human motion tracking applications.

Conclusions:

• The proposed quaternion-based complementary filter offers a computationally efficient solution for wearable motion tracking.
• The dynamic fusion approach enhances accuracy by adapting to sensor vector deviations.
• The validated filter design holds promise for advancing real-time human motion analysis in various fields.