Gait analysis is a standard practice used by clinicians and researchers to identify abnormalities, examine disease progression, or assess the success of interventions. Traditionally, assessments were performed with visual inspection by a trained professional. However, with the recent breakthroughs in sensing technologies, there is a growing body of literature that uses features extracted from sensing data as inputs to machine learning methods. These models require a large representative sample of gait cycles labeled according to each category of interest (e.g. standard, anomalous) for model training. This paper provides a personalized, nonparametric statistical framework that can be used for detecting and interpreting gait changes in individuals while requiring only a small number of baseline gait cycles. This framework can be applied using the acceleration trajectory or features from a single Intertial Measurement Unit (IMU). The individualized framework does not require the gait cycles to be labeled and does not require the assumption that the observed patterns are consistent across subjects. The personalized framework is applied to gait cycles extracted from a material handling task that simulates moving heavy loads in a warehouse. Twelve subjects were monitored and significant changes in personalized gait patterns consistent with perceived exertion were observed. Further interpretation of the changes illustrates that participants exhibit individualized patterns in gait as they approach the fatigued state.

Repetitive task performance is a leading cause of musculoskeletal injuries among order-picking workers in warehouses. The repetition of lifting tasks increases the risk of back and shoulder injuries among these workers. While lifting in this industry is composed of loaded and unloaded picking and placing, the existing literature does not address the separate analysis of the biomechanics of the back and shoulder for these events. To that end, we investigated the kinematics of the back and shoulder movements of nine healthy male participants who performed three sessions of a simulated de/palletization task. Their back and shoulder kinematics were sensed using an optical motion capture system to determine the back inclination and shoulder flexion. Comparison of the kinematics between the first and last sessions indicated statistically significant changes in the timings, angles, coordination between the back and shoulder, and moments around the shoulder ($p<0.05$). The majority of the significant changes were observed during the loaded events, which confirms the importance of the separation of these events for biomechanical analysis. This finding suggests that focusing worker evaluation on the loaded periods can provide important information to detect kinematic changes that may affect musculoskeletal injury risk.

Human activity recognition has been extensively used for the classification of occupational tasks. Existing activity recognition approaches perform well when training and testing data follow an identical distribution. However, in the real world, this condition may be violated due to existing heterogeneities among training and testing data, which results in degradation of classification performance. This study aims to investigate the impact of four heterogeneity sources, cross-sensor, cross-subject, joint cross-sensor and cross-subject, and cross-scenario heterogeneities, on classification performance. To that end, two experiments called separate task scenario and mixed task scenario were conducted to simulate tasks of electrical line workers under various heterogeneity sources. Furthermore, a support vector machine classifier equipped with domain adaptation was used to classify the tasks and benchmarked against a standard support vector machine baseline. Our results demonstrated that the support vector machine equipped with domain adaptation outperformed the baseline for cross-sensor, joint cross-subject and cross-sensor, and cross-subject cases, while the performance of support vector machine equipped with domain adaptation was not better than that of the baseline for cross-scenario case. Therefore, it is of great importance to investigate the impact of heterogeneity sources on classification performance and if needed, leverage domain adaptation methods to improve the performance.