Human foot force suggests different balance control between younger and older adults.

Aging can cause the decline of balance ability, which can lead to increased falls and decreased mobility. This work aimed to discern differences in balance control between healthy older and younger adults. Foot force data of 38 older and 65 younger participants (older and younger than 60 years, respectively) were analyzed. To first determine whether the two groups exhibited any differences, this study incorporated the orientation of the foot-ground interaction force in addition to its point of application. Specifically, the frequency-dependence of the "intersection point" of the lines of actions of the foot-ground interaction forces were evaluated. Results demonstrated that, like the mean center-of-pressure speed, a traditionally-employed measure, the intersection-point analysis could distinguish between the two participant groups. Then, to further explore age-specific control strategies, simulations of standing balance were conducted. An optimal controller stabilized a double-inverted-pendulum model with torque-actuated ankle and hip joints corrupted with white noise. The experimental data were compared to the simulation results to identify the controller parameters that best described the human data. Older participants showed significantly more use of the ankle than hip compared to younger participants. Best-fit controller gains suggested increased preference for asymmetric inter-joint neural feedback, possibly to compensate for the effects of aging such as sarcopenia. These results underscore the advantages of the intersection-point analysis to quantify possible shifts in inter-joint control with age, thus highlighting its potential to be used as a balance assessment tool in research and clinical settings.

Frequency-domain patterns in foot-force line-of-action: an emergent property of standing balance control.

A recent line of work suggests that the net behavior of the foot-ground interaction force provides insight into quiet-standing-balance dynamics and control. Through human subject experiments, Boehm et al. found that the relative variations of the center of pressure and force orientation emerge as a distinct pattern in the frequency domain, termed the "intersection-point height." Subsequent empirical and simulation-based studies showed that different control strategies are reflected in the distribution of intersection-point height across frequency. To facilitate understanding of the strengths and limitations of the intersection-point height in describing the dynamics and control of standing, the present work establishes a spectral-based method that also enables derivation of a closed-form estimate of the intersection-point height from any linear model of quiet stance. This new method explained observations from prior work, including how the measure captures aspects of control and physiological noise. The analysis presented herein highlights the utility of the frequency-dependent foot-force dynamics in probing the balance controller and provides a tool for model development and validation to further our understanding of the neuromotor control of natural upright posture in humans.

Acute effects of caffeine supplementation on kinematics and kinetics of sprinting.

We investigated the acute effects of caffeine supplementation (6 mgï½¥kg-1 ) on 60-m sprint performance and underlying components with a step-to-step ground reaction force measurement in 13 male sprinters. After the first round sprint as a control, caffeine supplementation-induced improvement in 60-m sprint times (7.811 s at the first versus 7.648 s at the second round, 2.05%) were greater compared with the placebo condition (7.769 s at the first versus 7.768 s at the second round, 0.02%). Using average values for every four steps, in the caffeine condition, higher running speed (all six step groups), higher step frequency (5th-16th and 21st-24th step groups), shorter support time (all the step groups except for 13th-16th step) and shorter braking time (9th-24th step groups) were found. Regarding ground reaction forces variables, greater braking mean force (13th-19th step group), propulsive mean force (1st-12th and 17th-20th step groups), and effective vertical mean force (9th-12th step group) were found in the caffeine condition. For the block clearance phase at the sprint start, push-off and reaction times did not change, while higher total anteroposterior mean force, average horizontal external power, and ratio of force were found in the caffeine condition. These results indicate that, compared with placebo, acute caffeine supplementation improved sprint performance regardless of sprint sections during the entire acceleration phase from the start through increases in step frequency with decreases in support time. Moreover, acute caffeine supplementation promoted increases in the propulsive mean force, resulting in the improvement of sprint performance.

Neuromuscular fatigue and cognitive constraints independently modify lower extremity landing biomechanics in healthy and chronic ankle instability individuals.

The purpose was to determine the impact of both cognitive constraint and neuromuscular fatigue on landing biomechanics in healthy and chronic ankle instability (CAI) participants. Twenty-three male volunteers (13 Control and 10 CAI) performed a single-leg landing task before and immediately after a fatiguing exercise with and without cognitive constraints. Ground Reaction Force (GRF) and Time to Stabilization (TTS) were determined at landing in vertical, anteroposterior (ap) and mediolateral (ml) axes using a force plate. Three-dimensional movements of the hip, knee and ankle were recorded during landing using a motion capture system. Exercise-induced fatigue decreased ankle plantar flexion and inversion and increased knee flexion. Neuromuscular fatigue decreased vertical GRF and increased ml GRF and ap TTS. Cognitive constraint decreased ankle internal rotation and increased knee and hip flexion during the flight phase of landing. Cognitive constraint increased ml GRF and TTS in all three axes. No interaction between factors (group, fatigue, cognitive) were observed. Fatigue and cognitive constraint induced greater knee and hip flexion, revealing higher proximal control during landing. Ankle kinematic suggests a protective strategy in response to fatigue and cognitive constraints. Finally, these two constraints impair dynamic stability that could increase the risk of ankle sprain.

Biomechanics of step-off drop landings are affected by limb dominance and lead limb in task initiation.

This study examines the effects of limb dominance and lead limb in task initiation on the kinetics and kinematics of step-off drop landings. Nineteen male participants performed drop landings led by the dominant and non-dominant limbs at 45-cm and 60-cm drop heights. Ground reaction force (GRF) and lower body kinematic data were collected. Between-limb time differences at the initial ground contact were calculated to indicate temporal asymmetry. Statistical Parametric Mapping (SPM) was applied for waveform analysis while two-way repeated measures ANOVA was used for discrete parameters. SPM results revealed greater GRF and lesser ankle dorsiflexion in the lead limb compared to the trail limb in 3 out of 4 landing conditions. The dominant limb displayed a greater forefoot loading rate (45 cm: p=.009, ηp2 = 0.438; 60 cm: p=.035, ηp2 = 0.225) and greater ankle joint quasi-stiffness (45 cm: p < .001, ηp2 = 0.360; 60 cm: p < .001, ηp2 = 0.597) than the non-dominant limb. Not all 380 trials were lead-limb first landings, with a smaller between-limb time difference (p=.009, d = 0.60) at 60 cm (4.1 ± 2.3 ms) than 45 cm (5.6 ± 2.7 ms). In conclusion, the step-off drop landing is not an ideal protocol for examining bilateral asymmetry in lower limb biomechanics due to potential biases introduced by limb dominance and the step-off limb.

Calibrating Low-Cost Smart Insole Sensors with Recurrent Neural Networks for Accurate Prediction of Center of Pressure.

This paper proposes a scheme for predicting ground reaction force (GRF) and center of pressure (CoP) using low-cost FSR sensors. GRF and CoP data are commonly collected from smart insoles to analyze the wearer's gait and diagnose balance issues. This approach can be utilized to improve a user's rehabilitation process and enable customized treatment plans for patients with specific diseases, making it a useful technology in many fields. However, the conventional measuring equipment for directly monitoring GRF and CoP values, such as F-Scan, is expensive, posing a challenge to commercialization in the industry. To solve this problem, this paper proposes a technology to predict relevant indicators using only low-cost Force Sensing Resistor (FSR) sensors instead of expensive equipment. In this study, data were collected from subjects simultaneously wearing a low-cost FSR Sensor and an F-Scan device, and the relationship between the collected data sets was analyzed using supervised learning techniques. Using the proposed technique, an artificial neural network was constructed that can derive a predicted value close to the actual F-Scan values using only the data from the FSR Sensor. In this process, GRF and CoP were calculated using six virtual forces instead of the pressure value of the entire sole. It was verified through various simulations that it is possible to achieve an improved prediction accuracy of more than 30% when using the proposed technique compared to conventional prediction techniques.

Estimation of Kinetics Using IMUs to Monitor and Aid in Clinical Decision-Making during ACL Rehabilitation: A Systematic Review.

After an ACL injury, rehabilitation consists of multiple phases, and progress between these phases is guided by subjective visual assessments of activities such as running, hopping, jump landing, etc. Estimation of objective kinetic measures like knee joint moments and GRF during assessment can help physiotherapists gain insights on knee loading and tailor rehabilitation protocols. Conventional methods deployed to estimate kinetics require complex, expensive systems and are limited to laboratory settings. Alternatively, multiple algorithms have been proposed in the literature to estimate kinetics from kinematics measured using only IMUs. However, the knowledge about their accuracy and generalizability for patient populations is still limited. Therefore, this article aims to identify the available algorithms for the estimation of kinetic parameters using kinematics measured only from IMUs and to evaluate their applicability in ACL rehabilitation through a comprehensive systematic review. The papers identified through the search were categorized based on the modelling techniques and kinetic parameters of interest, and subsequently compared based on the accuracies achieved and applicability for ACL patients during rehabilitation. IMUs have exhibited potential in estimating kinetic parameters with good accuracy, particularly for sagittal movements in healthy cohorts. However, several shortcomings were identified and future directions for improvement have been proposed, including extension of proposed algorithms to accommodate multiplanar movements and validation of the proposed techniques in diverse patient populations and in particular the ACL population.

Continuous Gait Phase Estimation for Multi-Locomotion Tasks Using Ground Reaction Force Data.

Existing studies on gait phase estimation generally involve walking experiments using inertial measurement units under limited walking conditions (WCs). In this study, a gait phase estimation algorithm is proposed that uses data from force sensing resistors (FSRs) and a Bi-LSTM model. The proposed algorithm estimates gait phases in real time under various WCs, e.g., walking on paved/unpaved roads, ascending and descending stairs, and ascending or descending on ramps. The performance of the proposed algorithm is evaluated by performing walking experiments on ten healthy adult participants. An average gait estimation accuracy exceeding 90% is observed with a small error (root mean square error = 0.794, R2 score = 0.906) across various WCs. These results demonstrate the wide applicability of the proposed gait phase estimation algorithm using various insole devices, e.g., in walking aid control, gait disturbance diagnosis in daily life, and motor ability analysis.

How to walk to reduce footstep noise in multi-story residential buildings.

Loud footsteps from upstairs cause disturbance to downstairs neighbours in multi-story residential buildings. In this experiment, we examined how participants walk when asked to walk quietly and evaluated the efficiency of their quiet walking patterns. Changes in vertical impact loading rates during the early stance phase, walking speed, and lower limb muscle activity when asked to walk quietly were evaluated from twenty-six young participants. Study data show that participants who struck the ground with the rearfoot reduced the impact loading rate by 44.6% with 29.3% slower walking speed than normal walking. Those who struck with the fore- or mid-foot reduced the impact loading by 69.2% with a 23.4% decrease in speed. Quiet walking with the non-rearfoot strike pattern reduced the impact loading by 48.7%, even when asked to walk as fast as normal walking. The results support the non-rearfoot strike pattern as an efficient walking strategy for lowering footstep impact.Practitioner summary: Data of this study show that voluntary gait alteration, such as adopting a non-rearfoot strike pattern, can reduce footstep impact. The study results propose that implementing such changes could be beneficial in addressing floor noise issues of multi-story residential buildings.Abbreviations: RFS: Rearfoot strike; NRFS: non-rearfoot strike; COP: Center of pressure; NW: Normal walking; QWs: Quiet walking at a preferred slower speed; QWn: Quiet walking at the speed of normal walking; EMG: Electromyography; BW: Body weight; iNEMG: integrated normalized EMG.

Spring-Mass Characteristics in Runners Before and After a 56-km Road Ultramarathon.

Ultramarathons are a unique model to study the effects of systemic fatigue in athletes. This investigation applied the spring-mass template to study runners before and 2 days after a road ultramarathon to characterize the effects of fatigue on systemic gait patterns. Overground kinetics were captured 7 days before and 2 days after the event in 14 runners. Traditional kinetic and spring-mass parameters were calculated, along with nonlinear regression-derived parameters and spring-mass model fit metrics. After the ultramarathon, vertical force magnitudes and loading rates were unchanged, but impact peaks increased (1.88 ± 0.08-1.95 ± 0.10 bodyweight). Ground contact times were modestly shorter (-3 ± 1 ms), resulting in increased leg stiffness (10.0 ± 0.5-10.3 ± 0.5 kN/m) with equivocal vertical stiffnesses. The deviation from the modeled spring-mass kinetics also increased (171.3 ± 15.0-181.4 ± 16.5 N). Overall, the systemic mechanical behaviors of the runners persisted despite the fatigue and stress induced by a road ultramarathon. These findings support previous observations that runners maintain gross mechanical behavior when fatigued with small compensatory changes in spatiotemporal and traditional spring-mass characteristics. However, these findings also suggest that the variability within that gross behavior may increase after stress, suggesting new opportunities for quantifying those deviations.

Association between physical performance during sit-to-stand motion and frailty in older adults with cardiometabolic diseases: a cross-sectional, longitudinal study.

BACKGROUND: Although physical performance tests of the lower extremities are used to assess sarcopenia and frailty, little is known about the mechanisms by which the parameters of ground reaction force (GRF) measured during sit-to-stand motion affect the frailty status in older adults. We aimed to examine the association between GRF parameters during sit-to-stand motion and the incidence of frailty in older adults. METHODS: This longitudinal study evaluated 319 outpatients aged ≥ 65 years with cardiometabolic diseases. The GRF parameters were measured using a motor function analyzer, in which the power, speed, and balance scores were calculated. Frailty was diagnosed using the modified version of the Cardiovascular Health Study (mCHS) and the Kihon Checklist (KCL). The independent associations between scores and frailty indices were assessed using multivariate binomial logistic regression analyses. Cox regression analysis was used to examine whether power and speed scores were associated with the incidence of frailty after adjusting for covariates. RESULTS: Logistic regression analyses adjusted for covariates showed that the power and speed scores were associated with frailty according to the mCHS criteria (power: OR = 0.37, 95% CI = 0.22-0.63; speed: OR = 0.64, 95% CI = 0.52-0.79) and KCL criteria (power: OR = 0.40, 95% CI = 0.26-0.62; speed: OR = 0.81, 95% CI = 0.69-0.96) at baseline. Receiver operating characteristic analyses revealed that the area under the curve values of power and speed scores for discriminating mCHS-defined frailty were 0.72 and 0.73. The Cox regression analysis showed that the speed score predicted the incidence of mCHS-defined (HR = 0.45, 95% CI = 0.22-0.92, P = 0.029) and KCL-defined (HR = 0.77, 95% CI = 0.60-0.99, P = 0.039) frailty, whereas the power score was associated with the incidence of KCL-defined frailty (HR = 0.72, 95% CI = 0.55-0.95, P = 0.02) after adjusting for covariates. CONCLUSIONS: The speed and power scores measured during sit-to-stand motion are predictive of frailty in older adults with cardiometabolic disease. Therefore, the GRF parameters measured during sit-to-stand motion could be an important indicator of frailty. Further studies are necessary to examine whether the GRF parameters can be improved by exercise or whether the changes in these parameters are associated with the improvement of frailty status.

Predicting vertical ground reaction force in rearfoot running: A wavelet neural network model and factor loading.

This study proposed a simple method for selecting input variables by factor loading and inputting these variables into a wavelet neural network (WNN) model to predict vertical ground reaction force (vGRF). The kinematic data and vGRF of 9 rearfoot strikers at 12, 14, and 16 km/h were collected using a motion capture system and an instrumented treadmill. The input variables were screened by factor loading and utilized to predict vGRF with the WNN. Nine kinematic variables were selected, corresponding to nine principal components, mainly focusing on the knee and ankle joints. The prediction results of vGRF were effective and accurate at different speeds, namely, the coefficient of multiple correlation (CMC) > 0.98 (0.984-0.988), the normalized root means square error (NRMSE) < 15% (9.34-11.51%). The NRMSEs of impact force (8.18-10.01%), active force (4.92-7.42%), and peak time (7.16-12.52%) were less than 15%. There was a small number (peak, 4.12-6.18%; time, 4.71-6.76%) exceeding the 95% confidence interval (CI) using the Bland-Altman method. The knee joint was the optimal location for estimating vGRF, followed by the ankle. There were high accuracy and agreement for predicting vGRF with the peak and peak time at 12, 14, and 16 km/h. Therefore, factor loading could be a valid method to screen kinematic variables in artificial neural networks.

Adolescents running in conventional running shoes have lower vertical instantaneous loading rates but greater asymmetry than running barefoot or in partial-minimal shoes.

Footwear may moderate the transiently heightened asymmetry in lower limb loading associated with peak growth in adolescence during running. This repeated-measures study compared the magnitude and symmetry of peak vertical ground reaction force and instantaneous loading rates (VILRs) in adolescents during barefoot and shod running. Ten adolescents (age, 10.6 ± 1.7 years) ran at self-selected speed (1.7 ± 0.3 m/s) on an instrumented treadmill under three counter-balanced conditions; barefoot and shod with partial-minimal and conventional running shoes. All participants were within one year of their estimated peak height velocity based on sex-specific regression equations. Foot-strike patterns, peak vertical ground reaction force and VILRs were recorded during 20 seconds of steady-state running. Symmetry of ground reaction forces was assessed using the symmetry index. Repeated-measures ANOVAs were used to compare conditions (α=.05). Adolescents used a rearfoot foot-strike pattern during barefoot and shod running. Use of conventional shoes resulted in a lower VILR (P < .05, dz = 0.9), but higher VILR asymmetry (P < .05) than running barefoot (dz = 1.5) or in partial-minimal shoes (dz = 1.6). Conventional running shoes result in a lower VILR than running unshod or in partial-minimal shoes but may have the unintended consequence of increasing VILR asymmetry. The findings may have implications for performance, musculoskeletal development and injury in adolescents.

Using Raw Accelerometer Data to Predict High-Impact Mechanical Loading.

The purpose of this study was to develop peak ground reaction force (pGRF) and peak loading rate (pLR) prediction equations for high-impact activities in adult subjects with a broad range of body masses, from normal weight to severe obesity. A total of 78 participants (27 males; 82.4 ± 20.6 kg) completed a series of trials involving jumps of different types and heights on force plates while wearing accelerometers at the ankle, lower back, and hip. Regression equations were developed to predict pGRF and pLR from accelerometry data. Leave-one-out cross-validation was used to calculate prediction accuracy and Bland-Altman plots. Body mass was a predictor in all models, along with peak acceleration in the pGRF models and peak acceleration rate in the pLR models. The equations to predict pGRF had a coefficient of determination (R2) of at least 0.83, and a mean absolute percentage error (MAPE) below 14.5%, while the R2 for the pLR prediction equations was at least 0.87 and the highest MAPE was 24.7%. Jumping pGRF can be accurately predicted through accelerometry data, enabling the continuous assessment of mechanical loading in clinical settings. The pLR prediction equations yielded a lower accuracy when compared to the pGRF equations.

Raw Acceleration from Wrist- and Hip-Worn Accelerometers Corresponds with Mechanical Loading in Children and Adolescents.

The purpose of this study was to investigate associations between peak magnitudes of raw acceleration (g) from wrist- and hip-worn accelerometers and ground reaction force (GRF) variables in a large sample of children and adolescents. A total of 269 participants (127 boys, 142 girls; age: 12.3 ± 2.0 yr) performed walking, running, jumping (<5 cm; >5 cm) and single-leg hopping on a force plate. A GENEActiv accelerometer was worn on the left wrist, and an Actigraph GT3X+ was worn on the right wrist and hip throughout. Mixed-effects linear regression was used to assess the relationships between peak magnitudes of raw acceleration and loading. Raw acceleration from both wrist and hip-worn accelerometers was strongly and significantly associated with loading (all p's < 0.05). Body mass and maturity status (pre/post-PHV) were also significantly associated with loading, whereas age, sex and height were not identified as significant predictors. The final models for the GENEActiv wrist, Actigraph wrist and Actigraph hip explained 81.1%, 81.9% and 79.9% of the variation in loading, respectively. This study demonstrates that wrist- and hip-worn accelerometers that output raw acceleration are appropriate for use to monitor the loading exerted on the skeleton and are able to detect short bursts of high-intensity activity that are pertinent to bone health.

Field-Based Biomechanical Assessment of the Snatch in Olympic Weightlifting Using Wearable In-Shoe Sensors and Videos-A Preliminary Report.

Traditionally, the biomechanical analysis of Olympic weightlifting movements required laboratory equipment such as force platforms and transducers, but such methods are difficult to implement in practice. This study developed a field-based method using wearable technology and videos for the biomechanical assessment of weightlifters. To demonstrate the practicality of our method, we collected kinetic and kinematic data on six Singapore National Olympic Weightlifters. The participants performed snatches at 80% to 90% of their competition one-repetition maximum, and the three best attempts were used for the analysis. They wore a pair of in-shoe force sensors loadsol® (novel, Munich, Germany) to measure the vertical ground reaction forces under each foot. Concurrently, a video camera recorded the barbell movement from the side. The kinematics (e.g., trajectories and velocities) of the barbell were extracted using a free video analysis software (Kinovea). The power-time history was calculated from the force and velocity data. The results showed differences in power, force, and barbell velocity with moderate to almost perfect reliability. Technical inconsistency in the barbell trajectories were also identified. In conclusion, this study presented a simple and practical approach to evaluating weightlifters using in-shoe wearable sensors and videos. Such information can be useful for monitoring progress, identifying errors, and guiding training plans for weightlifters.

Extended Application of Inertial Measurement Units in Biomechanics: From Activity Recognition to Force Estimation.

Abnormal posture or movement is generally the indicator of musculoskeletal injuries or diseases. Mechanical forces dominate the injury and recovery processes of musculoskeletal tissue. Using kinematic data collected from wearable sensors (notably IMUs) as input, activity recognition and musculoskeletal force (typically represented by ground reaction force, joint force/torque, and muscle activity/force) estimation approaches based on machine learning models have demonstrated their superior accuracy. The purpose of the present study is to summarize recent achievements in the application of IMUs in biomechanics, with an emphasis on activity recognition and mechanical force estimation. The methodology adopted in such applications, including data pre-processing, noise suppression, classification models, force/torque estimation models, and the corresponding application effects, are reviewed. The extent of the applications of IMUs in daily activity assessment, posture assessment, disease diagnosis, rehabilitation, and exoskeleton control strategy development are illustrated and discussed. More importantly, the technical feasibility and application opportunities of musculoskeletal force prediction using IMU-based wearable devices are indicated and highlighted. With the development and application of novel adaptive networks and deep learning models, the accurate estimation of musculoskeletal forces can become a research field worthy of further attention.

Novel 3D Force Sensors for a Cost-Effective 3D Force Plate for Biomechanical Analysis.

Three-dimensional force plates are important tools for biomechanics discovery and sports performance practice. However, currently, available 3D force plates lack portability and are often cost-prohibitive. To address this, a recently discovered 3D force sensor technology was used in the fabrication of a prototype force plate. Thirteen participants performed bodyweight and weighted lunges and squats on the prototype force plate and a standard 3D force plate positioned in series to compare forces measured by both force plates and validate the technology. For the lunges, there was excellent agreement between the experimental force plate and the standard force plate in the X-, Y-, and Z-axes (r = 0.950-0.999, p < 0.001). For the squats, there was excellent agreement between the force plates in the Z-axis (r = 0.996, p < 0.001). Across axes and movements, root mean square error (RMSE) ranged from 1.17% to 5.36% between force plates. Although the current prototype force plate is limited in sampling rate, the low RMSEs and extremely high agreement in peak forces provide confidence the novel force sensors have utility in constructing cost-effective and versatile use-case 3D force plates.

Validation of In-Shoe Force Sensors during Loaded Walking in Military Personnel.

The loadsol® wireless in-shoe force sensors can be useful for in-field measurements. However, its accuracy is unknown in the military context, whereby soldiers have to carry heavy loads and walk in military boots. The purpose of this study was to establish the validity of the loadsol® sensors in military personnel during loaded walking on flat, inclined and declined surfaces. Full-time Singapore Armed Forces (SAF) personnel (n = 8) walked on an instrumented treadmill on flat, 10° inclined, and 10° declined gradients while carrying heavy loads (25 kg and 35 kg). Normal ground reaction forces (GRF), perpendicular to the contact surface, were simultaneously measured using both the loadsol® sensors inserted in the military boots and the Bertec instrumented treadmill as the gold standard. A total of eight variables of interest were compared between loadsol® and treadmill, including four kinetic (impact peak force, active peak force, impulse, loading rate) and four spatiotemporal (stance time, stride time, cadence, step length) variables. Validity was assessed using Bland-Altman plots and 95% Limits of Agreement (LoA). Bias was calculated as the mean difference between the values obtained from loadsol® and the instrumented treadmill. Results showed similar force-time profiles between loadsol® sensors and the instrumented treadmill. The bias of most variables was generally low, with a narrow range of LoA. The high accuracy and good agreement with standard laboratory equipment suggest that the loadsol® system is a valid tool for measuring normal GRF during walking in military boots under heavy load carriage.

Prediction of Three-Directional Ground Reaction Forces during Walking Using a Shoe Sole Sensor System and Machine Learning.

We developed a shoe sole sensor system with four high-capacity, compact triaxial force sensors using a nitrogen added chromium strain-sensitive thin film mounted on the sole of a shoe. Walking experiments were performed, including straight walking and turning (side-step and cross-step turning), in six healthy young male participants and two healthy young female participants wearing the sole sensor system. A regression model to predict three-directional ground reaction forces (GRFs) from force sensor outputs was created using multiple linear regression and Gaussian process regression (GPR). The predicted GRF values were compared with the GRF values measured with a force plate. In the model trained on data from the straight walking and turning trials, the percent root-mean-square error (%RMSE) for predicting the GRFs in the anteroposterior and vertical directions was less than 15%, except for the GRF in the mediolateral direction. The model trained separately for straight walking, side-step turning, and cross-step turning showed a %RMSE of less than 15% in all directions in the GPR model, which is considered accurate for practical use.