Body-worn Inertial Measurement Units Research Articles

Uneven terrain affects metabolic cost and gait in simulated complex lunar surfaces

Objective. Upcoming missions of the National Aeronautics and Space Administration (NASA) to the Moon will include extensive human exploration of the lunar surface. Walking will be essential for many exploration tasks, and metabolic cost during ambulation on simulated complex lunar surfaces requires further characterization. In this study, ten healthy subjects (6 male and 4 female) participated in three simulated lunar terrain walking conditions at the NASA Johnson Space Center’s planetary ‘Rock Yard’: (1) flat terrain, (2) flat terrain with obstacles, and (3) mixed terrain. Approach. Energy expenditure and gait were quantified with a wearable metabolic energy expenditure monitoring system and body-worn inertial measurement units (IMUs), respectively. Main results. It was found that participants walking on the mixed terrain, representing the highest workload condition, required significantly higher metabolic costs than in other terrain conditions (p < 0.001). Additionally, our novel IMU-based gait variables discriminated different terrains and identified changes in gait in simulated lunar terrain environments. Significance. Our results showed that the various surface irregularities and inconsistencies could cause additional physical effort while walking on the complex terrain. These findings provide insight into the effects of terrain on metabolic energy expenditure during simulated lunar extravehicular activities.

The Effect of Cognitive Task Complexity on Healthy Gait in the Walking Corsi Test.

Dual-task activities are essential within everyday life, requiring visual-spatial memory (VSM) and mobility skills. Navigational memory is an important component of VSM needed to carry out everyday activities, but this is often not included in traditional tests such as the Corsi block tapping test (CBT). The Walking Corsi Test (WalCT) allows both VSM and navigational memory to be tested together, as well as allowing measures of gait to be collected, thus providing a more complete understanding of dual-task function. The aim of this study was to investigate the effect of an increasingly complex cognitive task on gait in a healthy adult population, using the WalCT and body-worn inertial measurement unit (IMU) sensors. Participants completed both the CBT and WalCT, where they were asked to replicate increasingly complex sequences until they were no longer able to carry this out correctly. IMU sensors were worn on the shins throughout the WalCT to assess changes in gait as task complexity increased. Results showed that there were significant differences in several gait parameters between completing a relatively simple cognitive task and completing a complex task. The type of memory used also appeared to have an impact on some gait variables. This indicates that even within a healthy population, gait is affected by cognitive task complexity, which may limit function in everyday dual-task activities.

Deep Learning for Laying Hen Activity Recognition Using Wearable Sensors

Laying hen activities in modern intensive housing systems can dramatically influence the policies needed for the optimal management of such systems. Intermittent monitoring of different behaviors during daytime cannot provide a good overview, since daily behaviors are not equally distributed over the day. This paper investigates the application of deep learning technology in the automatic recognition of laying hen behaviors equipped with body-worn inertial measurement unit (IMU) modules in poultry systems. Motivated by the human activity recognition literature, a sophisticated preprocessing method is tailored on the time-series data of IMU, transforming it into the form of so-called activity images to be recognized by the deep learning models. The diverse range of behaviors a laying hen can exhibit are categorized into three classes: low-, medium-, and high-intensity activities, and various recognition models are trained to recognize these behaviors in real-time. Several ablation studies are conducted to assess the efficacy and robustness of the developed models against variations and limitations common for an in situ practical implementation. Overall, the best trained model on the full-feature acquired data achieves a mean accuracy of almost 100%, where the whole process of inference by the model takes less than 30 milliseconds. The results suggest that the application of deep learning technology for activity recognition of individual hens has the potential to accurately aid successful management of modern poultry systems.

Evaluation of Error-State Kalman Filter Method for Estimating Human Lower-Limb Kinematics during Various Walking Gaits.

Inertial measurement units (IMUs) offer an attractive way to study human lower-limb kinematics without traditional laboratory constraints. We present an error-state Kalman filter method to estimate 3D joint angles, joint angle ranges of motion, stride length, and step width using data from an array of seven body-worn IMUs. Importantly, this paper contributes a novel joint axis measurement correction that reduces joint angle drift errors without assumptions of strict hinge-like joint behaviors of the hip and knee. We evaluate the method compared to two optical motion capture methods on twenty human subjects performing six different types of walking gait consisting of forward walking (at three speeds), backward walking, and lateral walking (left and right). For all gaits, RMS differences in joint angle estimates generally remain below 5 degrees for all three ankle joint angles and for flexion/extension and abduction/adduction of the hips and knees when compared to estimates from reflective markers on the IMUs. Additionally, mean RMS differences in estimated stride length and step width remain below 0.13 m for all gait types, except stride length during slow walking. This study confirms the method's potential for non-laboratory based gait analysis, motivating further evaluation with IMU-only measurements and pathological gaits.

Kick-scooters identification in the context of transportation mode detection using inertial sensors: Methods and accuracy

This work presents a novel transportation mode detection algorithm that handles the recognition of kick-scooters. In 2015, 10 minutes of data from a kick-scooter were considered in a transportation mode detection study, yielding a 56% F1-score. Since then, kick-scooters were not given much attention. Yet, kick-scooters are now very present in the urban transportation ecosystem, and their consideration in transportation studies has become a must. To fill this gap, 4 hours of kick-scooter signals were collected by 18 participants, with a set of 6 different kick-scooters, using 3 body-worn inertial measurement units. Obviously, kick-scooter patterns are classified in contrast with other modes of transportation. Two classification scenarios are considered in order to gradually increase the classification model complexity. The first scenario includes walking, biking, and kick-scooter, while the second considers public transport (tramway and bus) in addition to the former transportation modes. Results show that kick-scooters can be detected with an F1-score of 80% in the first scenario. Walking and public transport samples were still accurately classified in the second scenario, with an F1-score above 80% for both classes. However, bike and kick-scooter samples were both classified with lower F1-scores, equal to 59% and 64% respectively. Therefore, the main focus of future works should be directed toward the separability of kick-scooters and bikes when public transport is considered. The findings also suggest to place preferably the sensors in the trouser’s pocket, allowing for leg motion to be finely captured.

Longitudinal Monitoring of Progressive Supranuclear Palsy using Body-Worn Movement Sensors.

We have previously shown that wearable technology and machine learning techniques can accurately discriminate between progressive supranuclear palsy (PSP), Parkinson's disease, and healthy controls. To date these techniques have not been applied in longitudinal studies of disease progression in PSP. We aimed to establish whether data collected by a body-worn inertial measurement unit (IMU) network could predict clinical rating scale scores in PSP and whether it could be used to track disease progression. We studied gait and postural stability in 17 participants with PSP over five visits at 3-month intervals. Participants performed a 2-minute walk and an assessment of postural stability by standing for 30seconds with their eyes closed, while wearing an array of six IMUs. Thirty-two gait and posture features were identified, which progressed significantly with time. A simple linear regression model incorporating the three features with the clearest progression pattern was able to detect statistically significant progression 3 months in advance of the clinical scores. A more complex linear regression and a random forest approach did not improve on this. The reduced variability of the models, in comparison to clinical rating scales, allows a significant change in disease status from baseline to be observed at an earlier stage. The current study sheds light on the individual features that are important in tracking disease progression. © 2022 The Authors. Movement Disorders published by Wiley Periodicals LLC on behalf of International Parkinson and Movement Disorder Society.

Test-retest reliability of wireless inertial-sensor derived measurements of knee joint kinematics.

Advances in sensor technology have provided an opportunity to measure gait characteristics using body-worn inertial measurement units (IMUs). Whilst research investigating the validity of IMUs in reporting gait characteristics is extensive, research investigating the reliability of IMUs is limited. This study aimed to investigate the inter-session reliability of wireless IMU derived measures of gait (i.e., knee angle, range of motion) taking multiple test administrators into account. Fifteen healthy volunteers (43 ± 15 years) completed two visits. Within each visit, participants were required to perform two sets of 6 gait trials (6-metre walk tests). IMUs were placed on the participant in 7 locations on the lower limbs and waist. A different test administrator (n = 3) applied the IMUs at each set. At visit 2, this procedure was repeated with the same test administrators as visit 1. Kinematic measures of maximum angle (Knee_Max), minimum angle (Knee_Min), and range of motion (RoM) are reported for the left and right knee. The intraclass correlation coefficients (ICC), standard error of measurement (SEM) and minimum detectable change (MDC) are reported to determine IMU reliability. The results confirmed moderate to good inter-session reliability across all features (0.73-0.87). SEM values ranged from 1.21-3.32° and MDC values ranged from 3.37 - 9.21°. Therefore, IMUs appear to be a reliable method to determine inter-session gait characteristics across multiple test administrators.

Spatiotemporal parameters and gait variability in people with psoriatic arthritis (PsA): a cross-sectional study

BackgroundFoot involvement is a major manifestation of psoriatic arthritis (PsA) and can lead to severe levels of foot pain and disability and impaired functional mobility and quality of life. Gait spatiotemporal parameters (STPs) and gait variability, used as a clinical index of gait stability, have been associated with several adverse health outcomes, including risk of falling, functional decline, and mortality in a wide range of populations. Previous studies showed some alterations in STPs in people with PsA. However, gait variability and the relationships between STPs, gait variability and self-reported foot pain and disability have never been studied in these populations. Body-worn inertial measurement units (IMUs) are gaining interest in measuring gait parameters in clinical settings.ObjectivesTo assess STPs and gait variability in people with PsA using IMUs, to explore their relationship with self-reported foot pain and function and to investigate the feasibility of using IMUs to discriminate patient groups based on gait speed-critical values.MethodsTwenty-one participants with PsA (age: 53.9 ± 8.9 yrs.; median disease duration: 6 yrs) and 21 age- and sex-matched healthy participants (age 54.23 ± 9.3 yrs) were recruited. All the participants performed three 10-m walk test trials at their comfortable speed. STPs and gait variability were recorded and calculated using six body-worn IMUs and Mobility Lab software (APDM®). Foot pain and disability were assessed in participants with PsA using the foot function index (FFI).ResultsCadence, gait speed, stride length, and swing phase were significantly lower, while double support was significantly higher, in the PsA group (p < 0.006). Strong correlations between STPs and the FFI total score were demonstrated (|r| > 0.57, p < 0.006). Gait variability was significantly increased in the PsA group, but it was not correlated with foot pain or function (p < 0.006). Using the IMUs, three subgroups of participants with PsA with clinically meaningful differences in self-reported foot pain and disability were discriminated.ConclusionSTPs were significantly altered in participants with PsA, which could be associated with self-reported foot pain and disability. Future studies are required to confirm the increased gait variability highlighted in this study and its potential underlying causes. Using IMUs has been useful to objectively assess foot function in people with PsA.Trial registrationClinicalTrials.gov, NCT05075343, Retrospectively registered on 29 September 2021.

Gait Phase Estimation of Unsupervised Outdoors Walking Using IMUs and a Linear Regression Model

Human gait analysis and detection are critical for many applications, including wearable and rehabilitation robotic devices, reducing or tracking injury risk. The proposed work allows researchers to study the gait phase of human subjects in an unsupervised outdoor environment without the need for fixed thresholds and sensor-embedded insoles. We present an experimental protocol to label gait events based on patterns in human subjects from two body-worn inertial measurement units (IMUs). Gait patterns are developed using a force plate and a motion capture system. Upon defining the gait pattern, human subjects walk outdoors for forty minutes to train and test a principal component analysis (PCA)-based linear regression model. Next, gait phase estimation is performed using the defined patterns from other human subjects to accommodate cases where motion capture and force plate data are unavailable. Results showed a minimum normalized gait phase estimation error of 1.81 %, a maximum of 2.48 %, and an average of 2.21 ± 0.258 % for all subjects involved. Results are particularly significant because the proposed work can be expanded to precise control of human-assistive devices, rehabilitation devices, and clinical gait analysis.

Automatic Recognition and Analysis of Balance Activity in Community-Dwelling Older Adults: Algorithm Validation.

BackgroundClinical mobility and balance assessments identify older adults who have a high risk of falls in clinics. In the past two decades, sensors have been a popular supplement to mobility and balance assessment to provide quantitative information and a cost-effective solution in the community environment. Nonetheless, the current sensor-based balance assessment relies on manual observation or motion-specific features to identify motions of research interest.ObjectiveThe objective of this study was to develop an automatic motion data analytics framework using signal data collected from an inertial sensor for balance activity analysis in community-dwelling older adults.MethodsIn total, 59 community-dwelling older adults (19 males and 40 females; mean age = 81.86 years, SD 6.95 years) were recruited in this study. Data were collected using a body-worn inertial measurement unit (including an accelerometer and a gyroscope) at the L4 vertebra of each individual. After data preprocessing and motion detection via a convolutional long short-term memory (LSTM) neural network, a one-class support vector machine (SVM), linear discriminant analysis (LDA), and k-nearest neighborhood (k-NN) were adopted to classify high-risk individuals.ResultsThe framework developed in this study yielded mean accuracies of 87%, 86%, and 89% in detecting sit-to-stand, turning 360°, and stand-to-sit motions, respectively. The balance assessment classification showed accuracies of 90%, 92%, and 86% in classifying abnormal sit-to-stand, turning 360°, and stand-to-sit motions, respectively, using Tinetti Performance Oriented Mobility Assessment-Balance (POMA-B) criteria by the one-class SVM and k-NN.ConclusionsThe sensor-based approach presented in this study provided a time-effective manner with less human efforts to identify and preprocess the inertial signal and thus enabled an efficient balance assessment tool for medical professionals. In the long run, the approach may offer a flexible solution to relieve the community’s burden of continuous health monitoring.

Complex Deep Neural Networks from Large Scale Virtual IMU Data for Effective Human Activity Recognition Using Wearables

Supervised training of human activity recognition (HAR) systems based on body-worn inertial measurement units (IMUs) is often constrained by the typically rather small amounts of labeled sample data. Systems like IMUTube have been introduced that employ cross-modality transfer approaches to convert videos of activities of interest into virtual IMU data. We demonstrate for the first time how such large-scale virtual IMU datasets can be used to train HAR systems that are substantially more complex than the state-of-the-art. Complexity is thereby represented by the number of model parameters that can be trained robustly. Our models contain components that are dedicated to capture the essentials of IMU data as they are of relevance for activity recognition, which increased the number of trainable parameters by a factor of 1100 compared to state-of-the-art model architectures. We evaluate the new model architecture on the challenging task of analyzing free-weight gym exercises, specifically on classifying 13 dumbbell execises. We have collected around 41 h of virtual IMU data using IMUTube from exercise videos available from YouTube. The proposed model is trained with the large amount of virtual IMU data and calibrated with a mere 36 min of real IMU data. The trained model was evaluated on a real IMU dataset and we demonstrate the substantial performance improvements of 20% absolute F1 score compared to the state-of-the-art convolutional models in HAR.

Error-state Kalman filter for lower-limb kinematic estimation: Evaluation on a 3-body model

Human lower-limb kinematic measurements are critical for many applications including gait analysis, enhancing athletic performance, reducing or monitoring injury risk, augmenting warfighter performance, and monitoring elderly fall risk, among others. We present a new method to estimate lower-limb kinematics using an error-state Kalman filter that utilizes an array of body-worn inertial measurement units (IMUs) and four kinematic constraints. We evaluate the method on a simplified 3-body model of the lower limbs (pelvis and two legs) during walking using data from simulation and experiment. Evaluation on this 3-body model permits direct evaluation of the ErKF method without several confounding error sources from human subjects (e.g., soft tissue artefacts and determination of anatomical frames). RMS differences for the three estimated hip joint angles all remain below 0.2 degrees compared to simulation and 1.4 degrees compared to experimental optical motion capture (MOCAP). RMS differences for stride length and step width remain within 1% and 4%, respectively compared to simulation and 7% and 5%, respectively compared to experiment (MOCAP). The results are particularly important because they foretell future success in advancing this approach to more complex models for human movement. In particular, our future work aims to extend this approach to a 7-body model of the human lower limbs composed of the pelvis, thighs, shanks, and feet.

Kinematics-Based Control of an Inflatable Soft Wearable Robot for Assisting the Shoulder of Industrial Workers

Recent developments in soft active wearable robots can be used for upper extremity injury prevention for healthy industrial workers with better comfort than rigid systems, but there has not been control strategy proposals for such use cases. In this letter, we introduce a kinematics-based controller for an inflatable soft wearable robot that provides assistance to the shoulder quickly and accurately when needed during industrial use cases. Our approach is to use a state machine to classify user intent using shoulder and trunk kinematics estimated with body-worn inertial measurement units. We recruited eight participants to perform various tasks common in the workplace and assessed the controller's intent classification accuracy and response times, by using the users' reactions to cues as their ground truth intentions. On average, we found that the kinematics controller had 99% classification accuracy, and responded 0.8 seconds after the users reacted to the cue to begin work and 0.5 seconds after the users reacted to a cue to stop the task. In addition, we implemented an EMG-based controller for comparison, with state transitions determined by EMG-based thresholds instead of kinematics. Compared to the EMG controller, the kinematics controller required similar time to detect the users' intentions to stop overhead work but an additional 0.17 seconds on average for detecting users' intentions to begin. Although slightly slower, the kinematics controller still provided support prior to users' work initiations. We also implemented an online adaptive tuning algorithm for the kinematics controller to speed up response time while ensuring accuracy during offset transitions. This research paves the way for a further study of kinematics-based controller in a mobile system in real work environments.

Changes in prefrontal cortical activity and turning in response to dopaminergic and cholinergic therapy in Parkinson's disease: A randomized cross-over trial

IntroductionCholinergic dysfunction contributes to mobility deficits in Parkinson's disease (PD). People with PD rely on limited prefrontal executive-attentional resources for the control of locomotion, including turning. Cortical and behavioral responses to cholinergic augmentation during turning remains unclear. We examined prefrontal cortex (PFC) activity while turning-in-place and spatiotemporal measures of turns in response to usual dopaminergic medication and adjunct cholinergic augmentation. MethodsThis study consisted of a single-site, randomized, double-blind crossover trial. Twenty PD participants were assessed in the levodopa-off state and then randomized to either levodopa + donepezil (5 mg) or levodopa + placebo treatments for two weeks followed by a 2-week washout before crossover. The primary outcome was change from off state in PFC activity while turning-in-place (assessed with functional near-infrared spectroscopy). Secondary outcomes were changes in spatiotemporal turning measures (assessed with body-worn inertial measurement units) and accuracy in the secondary task. ResultsNineteen participants completed the trial. While levodopa + placebo had no effect on PFC activity when turning-in-place with a dual-task, levodopa + donepezil led to a large reduction in PFC activity (effect size, −0.82). Spatiotemporal measures of turning improved with both treatments, with slightly greater effect sizes observed for levodopa + donepezil. Additionally, the accuracy in the concurrent cognitive task improved only with levodopa + donepezil (effect size, 0.63). ConclusionThe addition of cholinergic therapy with donepezil (5 mg/day for 2 weeks) to standard dopaminergic therapy reduced the burden on prefrontal executive-attentional resources while turning with a dual-task and improved secondary task accuracy and turning.

The Instrumented Timed "Up & Go" Test Distinguishes Turning Characteristics in Vestibular Hypofunction.

Deficits in vestibular function increase the risk for falls while turning. However, the clinical assessment of turning in patients with vestibular dysfunction is lacking, and evidence is limited that identifies the effectiveness of vestibular physical therapy in improving turning performance. The purpose of this study was to quantify walking and turning performance during the instrumented Timed "Up & Go" (TUG) test using body-worn inertial measurement units (IMUs). Novel instrumented TUG parameters were investigated for ability to distinguish patients with unilateral vestibular deafferentation (UVD) from control groups and discriminate the differences in turning parameters of patients with UVD following vestibular physical therapy. Thirty-eight individuals were recruited following UVD surgery: 26 age-matched veteran controls with reports of dizziness not from a peripheral vestibular origin, and 12 age-matched healthy controls. Participants were donned with IMUs and given verbal instructions to complete the TUG test as fast as safely possible. The IMU-instrumented and automated assessment of the TUG test provided component-based TUG parameters, including the novel walk:turn ratio. Among the participants with UVD, 19 completed an additional instrumented TUG testing after vestibular physical therapy. The walk:turn time ratio showed that turning performance in patients with UVD before rehabilitation is significantly more impaired than both the individuals with nonperipheral conditions and healthy controls. Vestibular rehabilitation significantly improved turning performance and "normalized" their walk:turn time ratio compared with healthy controls. The duration of the straight walking component in individuals with UVD before vestibular physical therapy, however, was not significantly different compared with that component in people after vestibular physical therapy and in healthy controls. The IMU-instrumented TUG test can be used to distinguish individuals with vestibular deafferentation and to objectively quantify the change in their turning performance after vestibular physical therapy. The IMU-based instrumented TUG parameters have the potential to quantify the efficacy of vestibular physical therapy and be adopted in the clinic.

A Data-Driven Approach to Predict Fatigue in Exercise Based on Motion Data from Wearable Sensors or Force Plate

Fatigue increases the risk of injury during sports training and rehabilitation. Early detection of fatigue during exercises would help adapt the training in order to prevent over-training and injury. This study lays the foundation for a data-driven model to automatically predict the onset of fatigue and quantify consequent fatigue changes using a force plate (FP) or inertial measurement units (IMUs). The force plate and body-worn IMUs were used to capture movements associated with exercises (squats, high knee jacks, and corkscrew toe-touch) to estimate participant-specific fatigue levels in a continuous fashion using random forest (RF) regression and convolutional neural network (CNN) based regression models. Analysis of unseen data showed high correlation (up to 89%, 93%, and 94% for the squat, jack, and corkscrew exercises, respectively) between the predicted fatigue levels and self-reported fatigue levels. Predictions using force plate data achieved similar performance as those with IMU data; the best results in both cases were achieved with a convolutional neural network. The displacement of the center of pressure (COP) was found to be correlated with fatigue compared to other commonly used features of the force plate. Bland–Altman analysis also confirmed that the predicted fatigue levels were close to the true values. These results contribute to the field of human motion recognition by proposing a deep neural network model that can detect fairly small changes of motion data in a continuous process and quantify the movement. Based on the successful findings with three different exercises, the general nature of the methodology is potentially applicable to a variety of other forms of exercises, thereby contributing to the future adaptation of exercise programs and prevention of over-training and injury as a result of excessive fatigue.

Classification of runners’ performance levels with concurrent prediction of biomechanical parameters using data from inertial measurement units

Identification of runner’s performance level is critical to coaching, performance enhancement and injury prevention. Machine learning techniques have been developed to measure biomechanical parameters with body-worn inertial measurement unit (IMU) sensors. However, a robust method to classify runners is still unavailable. In this paper, we developed two models to classify running performance and predict biomechanical parameters of 30 subjects. We named the models RunNet-CNN and RunNet-MLP based on their architectures: convolutional neural network (CNN) and multilayer perceptron (MLP), respectively. In addition, we examined two validation approaches, subject-wise (leave-one-subject-out) and record-wise. RunNet-MLP classified runner’s performance levels with an overall accuracy of 97.1%. Our results also showed that RunNet-CNN outperformed RunNet-MLP and gradient boosting decision tree in predicting biomechanical parameters. RunNet-CNN showed good agreement (R2 > 0.9) with the ground-truth reference on biomechanical parameters. The prediction accuracy for the record-wise method was better than the subject-wise method regardless of biomechanical parameters or models. Our findings showed the viability of using IMUs to produce reliable prediction of runners’ performance levels and biomechanical parameters.

Accuracy of the Region of Limb Stability in Predicting Risk for Lower Limb Injury.

This study aimed to determine whether a measure of lower limb segment stability derived from body-worn inertial measurement units can predict risk for lower limb musculoskeletal injury in Division I Collegiate Football Players (D1CFP). The region of limb stability (ROLS) values were collected in a cohort of D1CFP during preseason. ROLS is a measure of knee joint stability, defined by thigh and shank excursion (cm) in the anterior-posterior and medial-lateral direction during single limb stance. The ROLS symmetry index (SI) (%) is the ratio between lower limb ROLS values where 100% suggests absolute symmetry. One-hundred and four D1CFP participated in this study and were divided into two groups: 1) no previous lower limb injury or no in-season injury (n = 70, "noninjured group") and 2) no previous lower limb injury, but in-season injury requiring surgery (n = 34, "injured group" group). The mean ± SD ROLS SI was 82.86% ± 14.75% and 65.58% ± 16.46% for the noninjured and injured group, respectively. Significant differences in ROLS SI were found between groups (P < 0.001). The ROLS SI demonstrated an area under the curve of 0.8 (P < 0.001; 95% confidence interval = 0.71-0.88) with an SE of 0.04, indicating that the ROLS SI has good predictive accuracy in detecting those healthy D1CFP at risk for lower limb injury resulting in surgery. The ROLS SI was found to have good predictive accuracy in detecting individuals at risk for injury that were healthy and asymptomatic during preseason testing. Increase in thigh and shank excursions and/or decrease in SI between lower limbs may be a predictor of risk for future injury.

Accurate Long-Term Multiple People Tracking using Video and Body-Worn IMUs.

Most modern approaches for video-based multiple people tracking rely on human appearance to exploit similarities between person detections. Consequently, tracking accuracy degrades if this kind of information is not discriminative or if people change apparel. In contrast, we present a method to fuse video information with additional motion signals from body-worn inertial measurement units (IMUs). In particular, we propose a neural network to relate person detections with IMU orientations, and formulate a graph labeling problem to obtain a tracking solution that is globally consistent with the video and inertial recordings. The fusion of visual and inertial cues provides several advantages. The association of detection boxes in the video and IMU devices is based on motion, which is independent of a person's outward appearance. Furthermore, inertial sensors provide motion information irrespective of visual occlusions. Hence, once detections in the video are associated with an IMU device, intermediate positions can be reconstructed from corresponding inertial sensor data, which would be unstable using video only. Since no dataset exists for this new setting, we release a dataset of challenging tracking sequences, containing video and IMU recordings together with ground-truth annotations. We evaluate our approach on our new dataset, achieving an average IDF1 score of 91.2%. The proposed method is applicable to any situation that allows one to equip people with inertial sensors.

Validity of Pendant-Based IMU Assessment of Postural Stability Under Varying Balance Conditions Compared to a Sensor Positioned on the Lower Back

Body-worn inertial measurement units (IMUs) have been widely used in postural stability and balance studies because of their low cost and high level of convenience. In most studies, single IMU sensors are put on the lower back attached to a belt, placing the sensor near the body’s center of mass (COM). For some populations, such as pregnant women, wearing the sensor on a belt over the lower back presents challenges in terms of fit and comfort. Thus, it may be necessary to identify a better location for the sensor and a more comfortable means for attaching the sensor to the body. This study aims to implement and test a novel pendant IMU sensor hanging from the subject’s neck and placed over the sternum. Three standing tasks (double-leg, tandem, and single-leg standing) were performed under open- and closed-eye conditions for preliminary assessments of the ability of the new sensor to discriminate between balance conditions. Standard deviations were analyzed in different conditions, along with ROC curves and ANOVA analysis. The results showed that the pendant sensor can detect the signals as good as the sensor on the waist.