Analyzing human lower limb kinematics during walking: a challenging study with minimal inertial sensor configuration

Wearable technology and sensor fusion techniques have enabled the study of mobility in real-life conditions, with a specific focus on Inertial Measurement Units (IMUs) due to their portability and versatility. A popular and powerful configuration proposed and validated in the literature (Salis, 2023) involves instrumenting the pelvis and feet with IMUs. This setup allows the capture of spatio-temporal parameters and center-of-mass movements but lacks information about joint angles. Consequently, the description of lower limb movements remains incomplete. To obtain a comprehensive understanding of joint angles and reconstruct lower limb kinematics, attaching an IMU to each segment forming the joint would be necessary. However, this approach increases experimental complexity, making it impractical for real-life scenarios. This thesis aimed to explore the feasibility of obtaining an acceptable description of joint kinematics during in-lab walking using this strategic three-IMU configuration, without escalating experimental complexity. To reconstruct the kinematic chain, we exploited both the spatio-temporal parameters and the orientations obtained from the IMUs, which are, however, affected by errors. For this reason, an optimization framework had to be implemented to fit these quantities to a biomechanical model of the lower limbs. The biomechanical model was based on the Denavit-Hartenberg convention and incorporated three rotations for each joint. These joints were defined according to the rotation sequences and axis orientations recommended by the ISB guidelines, which is useful for standardizing results. To limit the errors affecting the IMU-based position and orientation two set of constraints including limits on joint angles and gait-specific conditions were introducted. The Sequential Quadratic Programming algorithm was employed for optimization, which minimizes an objective function that considers the differences between the orientation and position quantities obtained from the model and those obtained from the three IMUs. The experimental validation, conducted in a controlled environment, compared the proposed IMU-based system with a stereophotogrammetric system (SP), validating its accuracy and feasibility in a straight-line walking condition at comfort speed. Root Mean Square Error (RMSE) values, calculated by comparing joint angles from the proposed method with those from the SP after mean value removal, were 4.3, 6.1, and 4.8 degrees for hip, knee, and ankle joints, respectively. Generally, joint kinematics estimation systems are deemed acceptable with RMSE errors below 5 degrees (McGinley, 2009). This study highlighted the effectiveness of a minimal IMU configuration in capturing vital kinematic data during walking, presenting solutions for wearable sensor network challenges. Preliminary validation of the developed framework showed a significant step in the incorporation of IMU technology for real-world human movement analysis and its potential in providing more accessible and efficient clinical gait assessments. With further comprehensive validation, including several pathological cases and complex exercises, this technology could contribute to improved diagnosis and monitoring of motor disorders.