ThesisPDF Available

Prediction of Ground Reaction Forces in Running from Wearable Instrumentation and Algorithmic Models, PhD Thesis


Abstract and Figures

Measurement of ground reaction force (GRF) in running provides a direct indication of the loads to which the body is subjected, at each foot-ground contact, and can provide an objective explanation for performance outcomes. Traditionally, the collection of three orthogonal component GRF data in running requires an athlete to complete a series of return loops along a laboratory based runway, within which a force platform is embedded in order to collect data from a discrete footfall. The major disadvantages associated with this GRF data collection methodology includes, the inability to assess multiple consecutive foot contacts and the fact that measurements are typically confined to the laboratory. The objective of this research was to investigate the potential for wearable instrumentation to be employed, in conjunction with artificial neural network (ANN) and multiple linear regression (MLR) models, for the estimation of GRF in middle distance running. A custom, wearable data acquisition system was developed to acquire in-shoe force (ISF) and centre of mass acceleration (CMA) data simultaneously. Matched data sets from wearable instrumentation (source data) and force plate (target data) records were collected from elite middle distance runners under controlled laboratory conditions for the purposes of ANN and MLR model development (MD) and model validation (MV). Using a range of source data groupings, including ISF and CMA in isolation as well as in combination, it was found that ISF data, employed separately, provided the highest ANN and MLR model prediction accuracy for all three components of GRF. In general terms, an intrasubject, running speed interpolation based prediction scheme along with the MLR model was found to provide the highest prediction accuracy for the vertical (Correlation Coefficient [CC]: 0.997-1.000, Mean Absolute Error [MAE]: 14.494-46.658N) and medio-lateral (CC: 0.875-0.974, MAE: 10.680-39.890N) components of GRF. Alternatively, under the same prediction scheme, the ANN model provided the most accurate predictions of the anterior-posterior (CC: 0.9740.992, MAE: 14.910-31.118N) component of GRF. The prediction accuracy of each component of GRF was found to be governed by the inherent signal variability, in which case the vertical and anterior-posterior components were more reliable and subsequently predicted significantly more accurately than the medio-lateral component of GRF. In order to achieve accurate GRF predictions, the wearable instrumentation and algorithmic models need to have sufficient sensitivity to capture this inherent GRF signal variability. Findings from this research provide a proof of concept for the prediction of GRF from wearable instrumentation in middle distance running. The emerging capability for obtaining continuous GRF records from wearable instrumentation has the potential to permit unprecedented quantification and analysis of training stress and competition demands.
Content may be subject to copyright.
A preview of the PDF is not available
ResearchGate has not been able to resolve any citations for this publication.
Full-text available
Ground reaction forces (GRF) are associated with bone hypertrophy; therefore, they are important to understanding physical activity's role in children's bone health. In this study, we examined the ability of accelerometry to predict vertical GRF in 40 children (mean age 8.6 yr) during slow walking, brisk walking, running, and jumping. Correlation coefficients between accelerometry-derived movement counts and GRF were moderate to high and significant during walking and running, but not during jumping. Given a large proportion of children's daily physical activity involves ambulation, accelerometry should be useful as a research method in bone-related research. However, this method underestimates GRF during jumping, an important physical activity for bone modeling and remodeling.
Conference Paper
Full-text available
Introduction Considerable debate exists over the normal variation in gait of children. This debate focuses on the parameters, if any, that change as a function of age and walking speed. Research has shown that variables such as cadence, stance phase, stride length, stride width, and walking velocity change with age (Keen, 1992). However, such variables may also differ as a result of anthropometrics and mood. Wootten (1990) maintained that within a given age group, individuals may exhibit gait patterns that are not just variations on a theme, but in fact are completely different patterns. As a result of such work, normal kinetics and kinematics are often represented in a small number of curves averaged by age, or as a vast array of curves which attempt to account for all possible sources of variation. In the first case, the curves may not reflect the true diversity in gait patterns, while in the second, the diversity of patterns results in too few observations per category. Statement of Clinical Significance To address these issues, attempts were made to predict normal gait using neural networks. Neural nets have the advantage of being able to account for the natural variability of data due to their generalizability. Likewise, the neural net is a compact way to predict normal data without the use of an extensive look-up table or database. Success with this approach would also allow for the generation of kinematics or kinetics for a normal subject from variables that include walking speed, age, and anthropometric values. Methodology Kinematic and kinetic data from 95 normal children between the ages of 3 and 18 were collected at self-selected slow, medium and fast walking speeds. The data were averaged by subject within each speed and then grouped by age and speed to detect trends. Kinetic variables that exhibited trends by age or speed were noted for use in the neural networks. Kinematic variables were used in networks although no trends were apparent. Neural networks for each variable were created and trained to predict kinematics and kinetics based on subject age and forward walking velocity. Data for each gait cycle were compressed using five real and four imaginary FFT coefficients in order to more quickly and accurately train the neural networks. Ward backpropagation architecture which implemented the turboprop algorithm was used in the network due to its excellent predictive track record. Kinematic networks were run for hip, knee, and ankle flexion, and kinetic networks were run for ankle and hip flexion moments, hip abduction moments, ankle flexion moments, and vertical ground reaction forces. The data in each network were divided into training, test, and cross-validation sets. All statistical and graphical analyses of the network results were performed using only the cross-validation data sets.
Help your students understand the full continuum of human movement potential with the Fourth Edition of this rigorous-yet understandable-introductory text on the market. Focusing on the quantitative nature of biomechanics, Biomechanical Basis of Movement, Fourth Edition integrates current literature, meaningful numerical examples, relevant applications, hands-on exercises, and functional anatomy, physics, calculus, and physiology to help your students develop a holistic understanding of human movement. The book's chapters are essentially self-contained, allowing you maximum teaching flexibility in structuring your course. The Fourth Edition offers new content, new examples and applications, and online teaching and learning resources to save you time and help your students succeed. Instructor Resources: • NEW! A robust problem generator randomly generates an unlimited number of numerical problems you can assign to students for practice and self-testing. • Brownstone test generator, loaded with pre-made text-specific questions, saves you time and makes creating and printing tests easy. • Pre-loaded PowerPoint presentations speed lecture preparation. • A complete image bank enhances lecture and exam preparation. • WebCT and Blackboard Ready Cartridges allow you to connect to your preferred course management system with ease. Student Resources: • Answers to the text's review questions help students master key concepts. • Confidence-building practice quizzes allow students to test their understanding of key concepts and prepare for exams. • MaxTRAQ motion analysis software brings concepts to life and allows students to track data and analyze motion in a in a dynamic, video-enriched environment. • The fully searchable textbook online is ideal for review on the go! Handy online appendices present information on units of measurement and trigonometric functions, as well as hands- on data, for quick reference. © 2015, 2009, 2003, 1995 Lippincott Williams & Wilkins. All rights reserved.
The Paromed Datalogger® with two insole pressure transducers (16 sensors each, 200 Hz) was applied to study the feasibility of the system for measurement of plantar pressure distribution in ski jumping. The specific aim was to test the sensitivity of the Paromed system to the changes in plantar pressure distribution in ski jumping. Three international level ski jumpers served as subjects during the testing of the system. The Datalogger was fixed to the jumpers' lower back under the jumping suit. A separate pulse was transmitted to the Datalogger and tape recorder in order to synchronize the logger information with photocell signals indicating the location of the jumper on the inrun. Test procedure showed that this system could be used in ski jumping with only minor disturbance to the jumper. The measured relative pressure increase during the inrun curve matched well the calculated relative centrifugal force (mv2 · r-1), which thus serves a rough estimation of the system validity. Strong increase in pressure under the big toes compared to the heels (225% and 91%, respectively) with large interindividual differences characterized the take-off. These differences may reflect an unstable anteroposterior balance of a jumper while he tries to create a proper forward rotation for a good flight position.
The aim of our study was to assess the interday test-retest reliability (focussing on the separate contribution of systematic and random error) of selected 10-trial mean ground reaction force (GRF) parameters and GRF symmetry indices measured during running. Ten competitive male heel-strike runners (aged, 26.2 ± 5.7 years) performed 10 successful running trials across the force platform at a constant velocity of 4.0 m · s -1 ± 10% wearing their customary running footwear. The testing procedure was repeated under similar conditions 1 week later. The results showed no statistically significant differences between the means of Test 1 and Test 2 for most GRF parameters and symmetry indices, indicating non-significant systematic error. Correlation coefficients ranged from 0.73 to 0.99 for GRF parameters. Random error was small, with SE(meas) less than 10% of the Test 1 mean value for almost all GRF parameters. Symmetry indices displayed correlation coefficients ranging from -0.44 to 0.91. Based on these and the size of the SE(meas), the symmetry indices displayed variable reliability, with the most reliable being those associated with peak vertical active force and peak horizontal propulsive force.
The ADXL202 is the newest low-g (± 2 g), dual-axis, surface-micromachined accelerometer from Analog Devices. Building on experience gained in manufacturing millions of iMEMS ® accelerometers in the past six years, the ADXL202 is the world's first commercial dual axis, surface micromachined accelerometer to combine low-g sensing with lowest power, lowest noise, and digital outputs—all on a single silicon chip. Surface micromachining, first commercialized with the ADXL50, allows for integration of the acceleration sensor with all signal conditioning electronics—tight integration of the sensor and its signal conditioning is what has made this impressive performance possible. Lower cost was a major driver in the ADXL202 design effort. Integrating two axes resulted in a significant cost reduction per axis. In addition, while the ADXL50, ADXL150, and ADXL250 can be thought of as "acceleration to volts" transducers, the ADXL202 adds a Pulse Width Modulated (PWM) digital output capability as well. Since most accelerometers will interface with a microcontroller, a PWM output obviates the need for an A to D converter, further driving down the user's total system cost.