The dynamics of locomotion involve a fine-tuned, continuous feedback loop between processes in the brain, functioning of the muscles, and interactions with the environment. Neurological or motor disability can often disrupt this loop and alter muscle activation patterns and corresponding behavior. In order to maintain some level of function, the brain and body adopt atypical locomotive strategies that are often sub-optimal, which can have negative impacts on overall health and inhibit continued motor learning. Therefore, it is crucial to accurately and holistically characterize and diagnose motor behavior when providing interventions. In this dissertation, I propose an approach for comprehensively describing the variations in motor behavior within and across individuals, in addition to an approach for relating brain activity to motor behavior.

Many traditional locomotive analyses utilize subjective metrics derived from manually observed measures of the behavior, making it infeasible for high volumes of data with large numbers of trials or individuals. Additionally, broad summary statistics from a subset of the behavioral measures are typically used but fall short of capturing the full context or the inter-dependencies of the most important locomotive variables. In this work, I develop an approach to uncover the underlying characteristics defining distinct locomotive behaviors across multiple limbs and individuals, simultaneously. I apply higher-order statistics, not utilized in standard gait analyses, to identify time-varying, multi-modal activation patterns for comprehensive descriptions and visualizations. With these methods, I describe muscle recruitment strategies during gait of individuals with and without osteomyoplastic transfemoral amputation (OTFA) using pressure and electromyography (EMG) data and provide a robust approach for extracting, characterizing, and grouping the motor behavior across strides. I demonstrate the presence of muscle activity within the distal-residuum of multiple individuals with OTFA, which has not been shown before. I provide a novel perspective on co-contraction and compare the distributions of co-contraction timing between individuals with and without OTFA. Additionally, I provide quantitative descriptions of the distribution of pressures to objectively determine the quality of prosthetic fit. These results have potential implications for improving rehabilitation outcomes, prosthetic design, and reducing the risk of injury.

I also propose an approach for relating limb movements (from kinematic data) with brain activity (from electroencephalography, EEG) in infants during the acquisition of crawling. In this approach, I decompose the EEG signals into constituent frequency components and measure their relevance using machine learning models that predict movement from three developmental periods. This approach enables the examination of longitudinal changes in brain activity as infants are learning to crawl. I demonstrate that multiple frequency components of the EEG signals at distinct locations are relevant for predicting limb movements and I provide evidence for increasing functional connectivity at higher brain activation frequencies.