Austin R. Hess

Estimating load carriage from a body-worn accelerometer

Heavy loads increase the risk of musculoskeletal injury for foot soldiers and first responders. Continuous monitoring of load carriage in the field has proven difficult. We propose an algorithm for estimating load from a single body-worn accelerometer. The algorithm utilizes three different methods for characterizing torso movement dynamics, and maps the extracted dynamics features to load estimates using two machine learning multivariate regression techniques. The algorithm is applied, using leave-one-subject-out cross-validation, to two field collections of soldiers and civilians walking with varying loads. Rapid, accurate estimates of load are obtained, demonstrating robustness to changes in equipment configuration, walking conditions, and walking speeds. On soldier data with loads ranging from 45 to 89 lbs, load estimates result in mean absolute error (MAE) of 6.64 lbs and correlation of r = 0.81. On combined soldier and civilian data, with loads ranging from 0 to 89 lbs, results are MAE = 9.57 lbs and r = 0.91.

Individualized detection of ambulatory distress in the field using wearable sensors

The early onset of musculoskeletal injury during ambulation may be detectable due to changes in gait. Body worn accelerometers provide the ability for real-time monitoring and detection of these changes, thereby providing a means for avoiding further injury. We propose algorithms for extracting magnitude and pattern asymmetry features from accelerometers attached to each foot. By registering simultaneous acceleration differences between the two feet, these features provide robustness to a variety of confounding factors, such as changes in walking speed and load carriage. By computing only summary statistics from the acceleration signals, the algorithms can be easily implemented in real-time physiological status monitoring systems. We evaluate the algorithms on a field collection consisting of 32 subjects completing a series of 5 km marches under different loading conditions. We show that changes in the magnitude and pattern asymmetry features are predictive of subject ratings of physical pain and discomfort.

Predicting sound-localization performance with hearing-protection devices using computational auditory models

Evaluation of the effect of hearing-protection devices (HPDs) on auditory tasks such as detection, localization, and speech intelligibility typically is done with human-subject testing. However, such data collections can be impractical due to the time-consuming processes of subject recruitment and the testing itself, particularly when multiple tasks and HPDs are included. An alternative, objective testing protocol involves the use of a binaural mannequin (a.k.a an acoustic test fixture) and computational models of the auditory system. For example, data collected at the eardrums of such a mannequin outfitted with an HPD can be fed into a binaural localization model. If the performance of the model with such input can be shown to be similar to that of human subjects, the model-based assessment may be sufficient to characterize the hearing protector and inform further design decisions. In this presentation we will describe the preliminary results of an effort to replicate human-subject localization performan...

Using collar-worn sensors to forecast thermal strain in military working dogs

Military working dogs (MWDs) are at high risk of heat strain both during training and missions. Body heat in a MWD increases due to work, and the primary means for reducing this heat are resting and panting. Body-worn sensors can enable monitoring of work level and respiratory rate in real time. They can thereby provide real-time objective indicators of thermal strain in MWDs. In this paper a system is proposed for using collar-worn accelerometer, global positioning system (GPS), and audio recorder sensors to provide real-time estimates of work level and respiration (breathing and panting) rate. Automated methods are demonstrated for using a collar-worn accelerometer and GPS sensor to estimate work levels during multiple short-duration activities, and for estimating respiration rates from a collar-worn audio recorder. The potential utility of these estimates for forecasting and monitoring thermal strain is assessed based on performance in out of sample prediction of core temperature (Tc) statistics, which are obtained from ingestible sensors. Using cross-validation, regression models are trained from accelerometer- and GPS-based activity estimates to predict rate of change in Tc, obtaining a correlation of r=0.59 between actual and predicted Tc change rates. Regression models are also trained from audio-based respiration rate estimates during recovery to predict the Tc values immediately prior to recovery, obtaining a correlation of r=0.49 between actual and predicted Tc.

Detecting and tracking gait asymmetries with wearable accelerometers

Gait asymmetry can be a useful indicator of a variety of medical and pathological conditions, including musculoskeletal injury (MSI), neurological damage associated with stroke or head trauma, and a variety of age-related disorders. Body-worn accelerometers can enable real-time monitoring and detection of changes in gait asymmetry, thereby informing medical conditions and triggering timely interventions. We propose a practical and robust algorithm for detecting gait asymmetry based on summary statistics extracted from accelerometers attached to each foot. By registering simultaneous acceleration differences between the two feet, these asymmetry features provide robustness to a variety of confounding factors, such as changes in walking speed and load carriage. Evaluating the algorithm on natural walking data with induced gait asymmetries, we demonstrate that the extracted features are sensitive to the sign and magnitude of gait asymmetries and enable the detection and tracking of asymmetries during continuous monitoring.

Building animats: neurobiomimetic approach for cognitive systems

This paper discusses selected aspects of our effort to implement rudimentary emulations of brain regions and their synaptic connectivity (a paradigm we refer to as neurobiomimetic), including in particular the Neurobiomimetic Cognitive Architecture Foundation Framework (NCAFF) we developed. Different instances of neurobiomimetic emulations are possible and we refer to those instances as animats. NCAFF addresses the overwhelming developmental challenge of implementing animats, separating animat-development process from the lower-level details. Approaches such as NCAFF enable feasible building of animats which, by emulating aspects of brain structure and processes, may constitute a particularly promising path to cognitive systems.