Human gait monitoring: methods and systems using wearable technologies

Abstract

Several diseases and accidents can lead to motor impairments, preventing humans from nor mal daily life activities. In order to diagnose and treat the population suffering from walking disabilities, clinicians and physical therapists need tools that help to assess and analyze gait patterns. Nowadays, the gold standard in motion assessment are systems comprised by infrared high speed cameras and reflective markers. However, such systems are expensive and require a dedicated environment, limiting their use to indoors ambients and constrained spaces. Al ternatively, new sensor approaches are now shifting the paradigm from the bulk and expensive systems to wearable and more affordable technologies. Among others, inertial measurement units (IMU) are being widely used to assess human movements with little interference to user activities. Moreover, recent studies have demonstrated the feasibility of using optical fiber based curvature sensors to measure joint angles. Their adaptability, low-cost, light-weight and electromagnetic immunity are features that make them an interesting alternative technology. As a first contribution of this Ph.D thesis, we present a novel calibration procedure as a method to align IMUs to body segments, which, compared to other methods in the literature, is a faster and simpler sensor placement method, with no need predefined movements at calibration nor any additional tools. The promising results demonstrate the potential of this IMU-to-body alignment method to become an alternative to high-cost camera-based systems, allowing the possibility of performing human gait analysis in external environments, and with clinical ap plication in the near future. As a second contribution, we developed a novel IMU-POF sensor fusion system for knee angle monitoring, which consists of merging signals from two IMUs and a polymeric optical fiber (POF) curvature sensor. The fusion method relies more on IMUs or POF curvature sensor data depending on the gait cycle phase, generating a filtered output that is more accurate than any of the independent sensors. Our proposed system presented better performance (mean RMSE < 3.3 , LFM coefficients a1 = 0.99 0.04, a0 = 0.70 2.29, R2 = 0.98 0.01 and C >0.99) when compared to other methods in the literature. In summary, this Ph.D. thesis contributes to the state-of-the-art about the use of wearable technologies for mo tion analysis by improving the accuracy and usability of new sensors towards in-home motion monitoring and clinical scenarios