A Joint-Level Hybrid Framework for Gait Analysis Using Camera-IMU Fusion and LSTM-Based Temporal Correction

Area of Science:

• Biomechanics
• Human Movement Analysis
• Wearable Technology

Background:

• Marker-based motion capture is costly and spatially restrictive.
• Accurate gait analysis is crucial for diagnosing musculoskeletal disorders and rehabilitation.
• Existing markerless systems face challenges in precision and portability.

Purpose of the Study:

• To develop and validate a hybrid framework for joint-level gait analysis using a single RGB camera and two shoe-mounted inertial measurement units (IMUs).
• To overcome the limitations of traditional marker-based systems by integrating complementary sensor strengths.
• To provide a practical and interpretable approach for portable lower limb gait analysis.

Main Methods:

• A hybrid framework combining a single RGB camera for hip and knee sagittal angle estimation (using 3D pose estimation, DEAS optimization, and LSTM refinement) and two IMUs for ankle angle estimation.
• Utilized kinematic chain relationships, integrating camera-derived proximal joint data with IMU-measured foot orientation.
• Employed leave-one-subject-out (LOSO) cross-validation for the LSTM correction model to assess generalizability.

Main Results:

• Achieved promising estimation performance with an average Mean Absolute Error (MAE) of 7.89° and Root Mean Square Error (RMSE) of 10.09° for sagittal hip, knee, and ankle angles on a held-out test set.
• LOSO cross-validation of the LSTM correction model demonstrated generalizability, yielding an average MAE of 6.40° for bilateral hip angles.
• Successfully mitigated trunk-inclination-induced overestimation of hip angles.

Conclusions:

• The proposed hybrid framework offers a practical and interpretable solution for portable lower limb gait analysis.
• The integration of a single camera and two IMUs provides accurate joint-level angle estimations.
• This approach overcomes the cost and spatial limitations of traditional marker-based systems, enhancing clinical applicability.