Continuous Classification of Locomotion in Response to Task Complexity and Parkinson's Disease

Intent recognition in lower-extremity assistive devices (e.g., prostheses and exoskeletons) is typically limited to either recognition of steady-state locomotion or changes of terrain (e.g., level ground to stair) occurring in a straight-line path and under anticipated cognitive state. Stability is highly affected during non-steady changes of direction such as cuts especially when they are unanticipated, posing high risk of fall-related injuries for individuals with physical or/and cognitive impairments. Here, we studied the influence of changes of direction and user anticipation on task recognition, and accordingly introduced classification schemes accommodating such effects. A linear discriminant analysis (LDA) classifier continuously classified straight-line walking, sidestep/crossover cuts (single transitions), and cuts-to-stair locomotion (mixed transitions) performed under varied subjects anticipatory conditions. Training paradigms with varying levels of anticipated/unanticipated exposures as well as analysis windows of size 100-600 ms were examined. Relatively accurate (85-97%) recognition of anticipated tasks was reported. However, including bouts of target task in the training data was necessary to improve generalization to unanticipated locomotion. Window size modifications did not have a significant influence on classification performance. In the second aim of this study, the performance of accelerographic, gyroscopic, joint angle signals, and their fusion were compared across unilateral segments and torso to identify the most critical information during straight-walking, changes of direction (cuts), and cuts to stair ascent performed under varying user anticipatory states. Biomechanical information from the torso demonstrated poor task generalization. Our results suggest that regardless of whether an assistive device is on the leading or training leg, contralateral signals could be sufficient to provide an accurate recognition. Joint angles appeared to provide comparable performance to fusion of two/three signal modalities, and fusion of unilateral and torso data did not outperform unilateral signals only, suggesting that recognition accuracy may not increase as more sensors are added. Diminished accuracy rates were reported during unanticipated locomotion relative to anticipated (15-25%). Further, the dissertation attempts to address some limitations of existing activity recognition systems for individuals with motor impairments such as Parkinson’s disease. Fundamental knowledge in this area has been primarily limited to detection of steady-state/static tasks (e.g., sitting, standing, and walking). Neither identification of non-steady-state locomotion on uneven terrains (stairs, ramps) nor continuous tracking of these activities have received much attention. Furthermore, previous research has primarily relied on instrumenting a large number of body segments, which may be inconvenient for the user and could adversely affect system performance. Here, optical motion capture data were collected from five individuals with PD and five healthy subjects. Participants performed non-steady-state circuit trials comprising stairs, ramp, and changes of direction. An offline analysis using linear discriminant analysis (LDA) and a LongShort Term Memory (LSTM) neural network was performed for task recognition. The performance of accelerographic and gyroscopic information from varied lower/upper-body segments were tested across a set of user-independent and user-dependent training paradigms. Improved performance using LSTM compared to LDA was observed. Employing a mathematically simple classification algorithm such as LDA was shown to be at the expense of being limited to using a subject-dependent paradigm and/or instrumenting multiple body locations. The findings may inform classification schemes capable of adapting to complex locomotor activities, and could have implications for a number of applications in the field of healthcare monitoring by providing insights into classification schemes capable of handling non-steady-state and unstructured locomotion in individuals with mild Parkinson’s disease.