Accelerometer Vector Research Articles

Goats on the Move: Evaluating Machine Learning Models for Goat Activity Analysis Using Accelerometer Data.

Putting sensors on the bodies of animals to automate animal activity recognition and gain insight into their behaviors can help improve their living conditions. Although previous hard-coded algorithms failed to classify complex time series obtained from accelerometer data, recent advances in deep learning have improved the task of animal activity recognition for the better. However, a comparative analysis of the generalizing capabilities of various models in combination with different input types has yet to be addressed. This study experimented with two techniques for transforming the segmented accelerometer data to make them more orientation-independent. The methods included calculating the magnitude of the three-axis accelerometer vector and calculating the Discrete Fourier Transform for both sets of three-axis data as the vector magnitude. Three different deep learning models were trained on this data: a Multilayer Perceptron, a Convolutional Neural Network, and an ensemble merging both called a hybrid Convolutional Neural Network. Besides mixed cross-validation, every model and input type combination was assessed on a goat-wise leave-one-out cross-validation set to evaluate its generalizing capability. Using orientation-independent data transformations gave promising results. A hybrid Convolutional Neural Network with L2-norm as the input combined the higher classification accuracy of a Convolutional Neural Network with the lower standard deviation of a Multilayer Perceptron. Most of the misclassifications occurred for behaviors that display similar accelerometer traces and minority classes, which could be improved in future work by assembling larger and more balanced datasets.

Relationship between physical activity and central sensitization in chronic low back pain: Insights from machine learning

Background and objectivesChronic low back pain (CLBP) is a leading cause of disability. The management guidelines for the management of CLBP often recommend optimizing physical activity (PA). Among a subsample of patients with CLBP, central sensitization (CS) is present. However, knowledge about the association between PA intensity patterns, CLBP, and CS is limited. The objective PA computed by conventional approaches (e.g. cut-points) may not be sensitive enough to explore this association. This study aimed to investigate PA intensity patterns in patients with CLBP and low or high CS (CLBP-, CLBP+, respectively) by using advanced unsupervised machine learning approach, Hidden semi-Markov model (HSMM). MethodsForty-two patients were included (23 CLBP-, 19 CLBP+). CS-related symptoms (e.g. fatigue, sensitivity to light, psychological features) were assessed by a CS Inventory. Patients wore a standard 3D-accelerometer for one week and PA was recorded. The conventional cut-points approach was used to compute the time accumulation and distribution of PA intensity levels in a day. For the two groups, two HSMMs were developed to measure the temporal organization of and transition between hidden states (PA intensity levels), based on the accelerometer vector magnitude. ResultsBased on the conventional cut-points approach, no significant differences were found between CLBP- and CLBP+ groups (p = 0.87). In contrast, HSMMs revealed significant differences between the two groups. For the 5 identified hidden states (rest, sedentary, light PA, light locomotion, and moderate-vigorous PA), the CLBP- group had a higher transition probability from rest, light PA, and moderate-vigorous PA states to the sedentary state (p < 0.001). In addition, the CBLP- group had a significantly shorter bout duration of the sedentary state (p < 0.001). The CLBP+ group exhibited longer durations of active (p < 0.001) and inactive states (p = 0.037) and had higher transition probabilities between active states (p < 0.001). ConclusionsHSMM discloses the temporal organization and transitions of PA intensity levels based on accelerometer data, yielding valuable and detailed clinical information. The results imply that patients with CLBP- and CLBP+ have different PA intensity patterns. CLBP+ patients may adopt the distress-endurance response pattern with a prolonged bout duration of activity engagement.

Classroom Location, Activity Type, and Physical Activity During Preschool Children's Indoor Free-Play

Preschool children spend most of their day in the indoor classroom, but it is unknown which learning centers or activity types are associated with physical activity (PA) or if children even use learning centers for their intended purpose (e.g., block play in the block center). The present study aimed to describe activity types in which children participate within classroom locations and to compare location- versus activity type-based approaches to characterizing children's classroom PA. Preschoolers' (N = 50) classroom location and activity type were video-coded and matched with PA level (measured as accelerometer vector magnitude in counts/s) for one indoor free-play period. Multiple activities occurred within each location (median = 6.5), with the predominant activity type accounting for 40.7 (sociodramatic play in group location) to 88.4 (manipulative play in sensory location) percent of time in that location. Children spent the most time in the group location (17.9%) and the sociodramatic activity type (19.9%). PA varied by location and activity type with the loft location (37.3 counts/s) and gross motor activity type (46.3 counts/s) promoting the highest PA levels. Activity types elicited varying PA levels depending on location (e.g., block activity in block vs. group location: 22.8 vs. 43.9 counts/s). Children do not always participate in the type of activity for which a space was designed but this can facilitate PA. Activity type- and location-based approaches to characterizing children’s play and PA are not interchangeable but these approaches may provide additive information about classroom behavior, which is a methodological consideration for future studies.

Calibrating Wrist-Worn Accelerometers for Physical Activity Assessment in Preschoolers: Machine Learning Approaches.

BackgroundPhysical activity (PA) level is associated with multiple health benefits during early childhood. However, inconsistency in the methods for quantification of PA levels among preschoolers remains a problem.ObjectiveThis study aimed to develop PA intensity cut points for wrist-worn accelerometers by using machine learning (ML) approaches to assess PA in preschoolers.MethodsWrist- and hip-derived acceleration data were collected simultaneously from 34 preschoolers on 3 consecutive preschool days. Two supervised ML models, receiver operating characteristic curve (ROC) and ordinal logistic regression (OLR), and one unsupervised ML model, k-means cluster analysis, were applied to establish wrist-worn accelerometer vector magnitude (VM) cut points to classify accelerometer counts into sedentary behavior, light PA (LPA), moderate PA (MPA), and vigorous PA (VPA). Physical activity intensity levels identified by hip-worn accelerometer VM cut points were used as reference to train the supervised ML models. Vector magnitude counts were classified by intensity based on three newly established wrist methods and the hip reference to examine classification accuracy. Daily estimates of PA were compared to the hip-reference criterion.ResultsIn total, 3600 epochs with matched hip- and wrist-worn accelerometer VM counts were analyzed. All ML approaches performed differently on developing PA intensity cut points for wrist-worn accelerometers. Among the three ML models, k-means cluster analysis derived the following cut points: ≤2556 counts per minute (cpm) for sedentary behavior, 2557-7064 cpm for LPA, 7065-14532 cpm for MPA, and ≥14533 cpm for VPA; in addition, k-means cluster analysis had the highest classification accuracy, with more than 70% of the total epochs being classified into the correct PA categories, as examined by the hip reference. Additionally, k-means cut points exhibited the most accurate estimates on sedentary behavior, LPA, and VPA as the hip reference. None of the three wrist methods were able to accurately assess MPA.ConclusionsThis study demonstrates the potential of ML approaches in establishing cut points for wrist-worn accelerometers to assess PA in preschoolers. However, the findings from this study warrant additional validation studies.

Wheelchair Basketball Exercise Intensity in Youth

Children and youth with physical disabilities are at the greatest risk for inactivity, emphasizing the need for evidence-based activity options in recreation, rehabilitation, and post-rehabilitation settings. To determine if a wheelchair sport enabled youth participants to sustain moderate to vigorous physical activity (MVPA), the primary purpose of this study was to document exercise intensity of wheelchair basketball (WBB) among youth. The secondary purpose was to examine the influence of age, disability type, and training condition on intensity. Thirty-four youth with physical disabilities (MAge = 15.73 ± 2.57 yrs; MPlaying Experience = 4.32 ± 2.77 yrs) participated in three wheelchair basketball training sessions that included tactical drills (M = 113 min), and scrimmage (M = 57 min) training conditions. Percentage of maximum heart rate (HRMax), mean accelerometer vector counts per minute, and rate of perceived exertion (RPE) were assessed across 3 practices and collapsed into one training intensity score for each variable. Mean HRMax was 66±8% and mean accelerator vector counts were 6,055±1439 per minute, indicating that participants sustained MVPA during WBB training. Perceived effort, however, was distinct from the actual physical demand as mean RPE was 3.1±1.5. Age did not influence intensity scores but players who had a congenital or acquired spinal cord injury and were participating in scrimmage activity demonstrated higher exercise intensity. These findings support the ability of youth WBB players to sustain MVPA. This activity can be recommended as an evidence-based initiative in sport, rehabilitation or post-rehabilitation settings.

Particle motion measurements near a rocky shore off Cabo Frio Island.

This paper describes the lessons learned from the experiment BIOCOM'19 carried out in January 2019, in a shallow water bay off the island of Cabo Frio (RJ, Brazil). A dual accelerometer vector sensor hydrophone was deployed for two days, near a rocky shore covered with a significant benthic fauna. The results show that the frequency band above approximately 1.5 kHz is mostly associated with invertebrate biological noise and that the acoustic and the particle motion fields have a similar behavior, following the usual dawn-dusk activity pattern, and a coherent directivity content. At low frequencies, below ∼300 Hz, the acoustic pressure and the particle acceleration fields have significantly different spectral content along time. Many of these differences are due to anthropogenic noise sources related with nearby boating activity, while during quiet periods, they may be attributed to the biological activity from the rocky shore.

A validation study of the Eurostat harmonised European time use study (HETUS) diary using wearable technology

BackgroundThe central aim was to examine the accuracy of the full range of daily activities recorded in self-report time-use diaries against data from two objective passive data collection devices (wearable camera and accelerometer) serving as criterion reference instruments. This enabled systematic checks and comparisons on the timing, sequence and duration of activities recorded from the three data sources.MethodsParticipants (n = 148) were asked to complete a single-day self-report paper time-use diary designed for use in the Harmonised European Time Use Study (HETUS), while simultaneously wearing a camera that continuously recorded images of their activities, and an accelerometer tracking physical movement. In a reconstruction interview shortly after the data collection period, participants viewed the camera images to help researchers interpret the image sequences. Of the initial 148 recruits (multi-seed snowball sample, 59% women, aged 18–91, 43% > 40) 131 returned usable diary and camera records (of whom 124 also provided a usable whole-day accelerometer record. We compare time allocation estimates from the diary and camera records, and also match the diary and camera records to the simultaneously recorded accelerometer vector magnitudes.ResultsThe data were examined at three analytic levels: aggregate, individual diarist and timeslot. The most important finding is that the estimates of mean daily time devoted to 8 of the 10 main activities differ by < 10% in the camera and diary records. The single case of major divergence (eating) can be explained by a systematic difference between the procedures followed by the self-reporting diarist and the observer coding the camera records. There are more substantial differences at the respondent level, paired t-tests showing significant differences in time spent in the 4/10 categories. 45% of all variation in the accelerometer vector magnitudes in the timeslots is explained by camera and diary records. Detailed activity classifications perform much better than METs as predictors of actigraphy.ConclusionsThe comparison of the diary with the camera and accelerometer records strongly supports using diary methodology for studying the full range of daily activity, particularly at aggregate levels. Accelerometer data could be combined with diary measures to improve estimation of METs equivalents for various types of active and sedentary behaviour.

Device Measured Physical Activity And Sedentary Behavior In Chilean Older Adults: The Latin American Cities For Healthy Aging Study

PURPOSE: To describe device measured physical activity (PA) and sedentary behaviour (SB) in older adults from Temuco, Chile. METHODS: Community-dwelling adults aged 60 and older were recruited from 17 different neighbourhoods in Temuco, Chile. The participants were asked to wear an ActiGraph GT3X+ accelerometer on the hip for 7 consecutive days. A sleeping and a non-wear-time algorithms were used to detect daily wearing time. Accelerometer data were considered valid if the participant wore the device for at least 8 hours on any 4 days of the week. Accelerometer vector magnitude counts (cpm) were used to classify all worn time as either SB (<200 cpm), light (LPA, 200-2689 cpm), moderate (MPA, 2690-6166 cpm), vigorous (VPA, 6167-9642 cpm), very vigorous PA (VVPA, ≥9643 cpm) and moderate-to-vigorous PA (MVPA, ≥2690 cpm cpm) using a combination of Aguilar (2014) and Sasaki (2011) cut-points. For those participants who were unable to provide 7 complete days, total time spent in different activity intensities was calculated as: (5 × mean weekday time + 2 × mean weekend day time)/7, thus providing an average day estimate for each behavior. PA recommendation compliance was defined as accumulating either ≥ 150 min/week of MVPA or 30 min/day of MVPA on at least 5 days. Multivariate regressions were used for analysis. RESULTS: 316 participants were recruited, but 228 provided valid accelerometer data (69.5±6.86 years, 32% male). On an average day the participants spent 303.7±87.1 min/day in SB, 375.4±88.5 min/day in LPA and 40.8±29.3 min/day in MVPA. In relative terms, the older adults spent 42.3±11.5% of their waking hours in SB and only 5.6±3.9% in MVPA per day. Age was negatively associated with time spent in MPA (β=-0.75 min/day, p=0.003) and MVPA (β=-0.78 min/day, p<0.001). Being male was positively associated with MPA (β=14.7 min/day, p<0.001), VPA (β=1.0 min/day, p<0.001) and MVPA (β=15.7 min/day, p<0.001). 33.3% of the sample met the PA recommendations with no differences by gender. The oldest old were less likely to meet PA recommendations (OR=0.93, p=0.003). CONCLUSIONS: This study provides relevant information for reinforcing strategies to promote PA in elderly population, considering that the oldest old and women were less likely to engage in MVPA. Funded by CONICYT-CNPq/PCI/441970/2016-8 , DIUFRO16-0110 , DIE17-0006

The Relationship Between Intra-Abdominal Pressure and Body Acceleration During Exercise.

High intra-abdominal pressure (IAP) may influence the development of pelvic floor disorders. We and others have used intravaginal pressure transducers to measure IAP in women during exercise and daily activities, but utilizing the transducer for long-term measurements creates compliance issues. Waist-worn accelerometers are prominent in research and may be a reliable alternative for approximating IAP. We hypothesized that there are pair-wise positive correlations between the mean maximal accelerometer vector magnitude and 2 IAP measurements: mean maximal IAP and area under the curve (AUC). Twenty-five women who regularly participated in exercise performed 13 activities. Intra-abdominal pressure was measured with an intravaginal transducer and acceleration with a waist-worn accelerometer. We determined the mean maximal IAP, AUC for IAP, and mean maximal accelerometer vector magnitude for each activity and participant. The relationship between IAP and acceleration was determined by computing the Pearson correlation coefficient (R) and the 95% confidence interval for mean maximal accelerometer vector magnitude versus mean maximal IAP and mean maximal accelerometer vector magnitude versus AUC for IAP. The R values were 0.7353 for mean maximal accelerometer vector magnitude versus mean maximal IAP (including walking) and 0.5059 for mean maximal accelerometer vector magnitude versus AUC for IAP (excluding walking). Walking at 3 speeds, analyzed separately, presented R values of 0.72208 for mean maximal IAP and 0.21678 for AUC. Waist-worn accelerometers may provide a viable method for approximating mean maximal IAP in a population of women during most activities.

The effects of fatigue on the running profile of elite team sport athletes. A systematic review and meta-analysis.

With the underlying mechanisms of fatigue shown to be task-specific, gaining an understanding of changes within the running profile due to fatigue may provide a greater task-specific method for monitoring neuromuscular fatigue (NMF) in elite team sport athletes. This review aimed to identify, review and summarize the literature to understand the effects of fatigue on the running profile of elite team sport athletes. MEDLINE, SPORTDiscus, EMBASE and CINAHL were searched. To manage heterogeneity in athlete characteristics and sport demands, studies were limited to elite male field/court team sports, in which changes in running, gait or stride characteristics were assessed. Thirteen studies were included in the analysis with changes in sprint time, kinematic variables; contact time, center of mass vertical displacement and vertical stiffness, GPS accelerometer vector variables; z and y, and mechanical parameters; velocity and V0, all shown to be sensitive measures of fatigue. Due to having at least three sets of data points, meta-analysis was run for sprint time, with results showing a significant increase immediately postexercise (effect size [ES]±95% confidence interval [CI]; 1.67±1.61, P=0.04), but no change 24 hours postexercise (ES±95% CI; 0.93±3.21, P=0.57). The findings indicate that fatigue has a negative impact on the running profile in elite team sport athletes, offering an alternative task-specific measure of monitoring NMF in this population. However, due to large discrepancies in protocols, future research is required to allow greater uniformity in variables and methods.

Development and Testing of a Dual Accelerometer Vector Sensor for AUV Acoustic Surveys.

This paper presents the design, manufacturing and testing of a Dual Accelerometer Vector Sensor (DAVS). The device was built within the activities of the WiMUST project, supported under the Horizon 2020 Framework Programme, which aims to improve the efficiency of the methodologies used to perform geophysical acoustic surveys at sea by the use of Autonomous Underwater Vehicles (AUVs). The DAVS has the potential to contribute to this aim in various ways, for example, owing to its spatial filtering capability, it may reduce the amount of post processing by discriminating the bottom from the surface reflections. Additionally, its compact size allows easier integration with AUVs and hence facilitates the vehicle manoeuvrability compared to the classical towed arrays. The present paper is focused on results related to acoustic wave azimuth estimation as an example of its spatial filtering capabilities. The DAVS device consists of two tri-axial accelerometers and one hydrophone moulded in one unit. Sensitivity and directionality of these three sensors were measured in a tank, whilst the direction estimation capabilities of the accelerometers paired with the hydrophone, forming a vector sensor, were evaluated on a Medusa Class AUV, which was sailing around a deployed sound source. Results of these measurements are presented in this paper.

Dual accelerometer vector sensor mounted on an autonomous underwater vehicle (AUV)—Experimental results

In seismic geo-acoustic exploration, the use of ships equipped with long streamers is of major concern due to the complexity of the operations. The European project WIMUST aims to improve the efficacy of actual geo-acoustic surveys through the use of AUVs towing short streamers. A Dual Accelerometer Vector Sensor (DAVS) was developed in order to complement the streamer’s data, allowing for the reduction of their size and facilitating the operation of the WiMUST distributed configuration. Each DAVS consists of two tri-axial accelerometers and one hydrophone aligned along a vertical axis. This configuration has the ability to cancel or significantly attenuate the direct and the surface reflection paths, which are undesirable in seismic imaging. Calibration tests with the DAVS have already been performed; this paper presents experimental results on the estimation of azimuthal directions when the DAVS is in motion. Signals in the 1-2kHz band were emitted by a source deployed in a shallow pond at 1.5m depth and acquired by the DAVS mounted on a MEDUSA class AUV, which was following a pre-programed path with a 0.26m/s nominal speed. The azimuth estimates are coherent with the MEDUSA trajectories even in curved paths where the thruster noise increases.

Nonlinear Observer for Tightly Integrated Inertial Navigation Aided by Pseudo-Range Measurements

A modular nonlinear observer for inertial navigation aided by pseudo-range measurements is designed and analyzed. The attitude observer is based on a recent nonlinear complementary filter that uses magnetometer and accelerometer vector measurements to correct the quaternion attitude estimate driven by gyro measurements, including gyro bias estimation. A tightly integrated translational motion observer is driven by accelerometer measurements, employs the attitude estimates, and makes corrections using the pseudo-range and range-rate measurements. It estimates position, range bias errors, velocity and specific force in an earth-fixed Cartesian coordinate frame, where the specific force estimate is used as a reference vector for the accelerometer measurements in the attitude observer. The exponential stability of the feedback interconnection of the two observers is analyzed and found to have a semiglobal region of attraction with respect to the attitude observer initialization and local region of attraction with respect to translational motion observer initialization. The latter is due to linearization of the range measurement equations that is underlying the selection of injection gains by solving a Riccati equation. In typical applications, the pseudo-range equations admit an explicit algebraic solution that can be easily computed and used to accurately initialize the position and velocity estimates. Hence, the limited region of attraction is not seen as a practical limitation of the approach for many applications. Advantages of the proposed nonlinear observer are low computational complexity and a solid theoretical foundation.

Wheelchair-mounted accelerometers for measurement of physical activity

Purpose: To evaluate the validity of a wheelchair frame-mounted accelerometer for the assessment of physical activity of wheelchair users. Method: Twelve collegiate wheelchair basketball players participated in this study. The study was conducted in a modern indoor gymnasium at a university in the USA. A randomized, crossover experimental design was used to investigate accelerometer output, participant heart rate, and distance travelled. Participants performed two trials of wheeling at a combination of two different effort levels (light and moderate: Prescribed using perceived exertion) and two different modes (continuous and stop-go). Results: Accelerometer vector magnitude activity counts (VM), heart rate (HR), and distance travelled were significantly different between light and moderate effort (p < 0.01). The continuous and stop-go wheeling modes were not significantly different (p = 0.37) for VM, but were significantly different (p < 0.01) for both HR and distance. Between-trial reliability of all data collapsed for the effort and mode combinations were: VM: r = 0.85; HR: r = 0.86; and distance: r = 0.96. Conclusion: A wheelchair frame-mounted accelerometer differentiated between perceptually-prescribed low and moderate effort levels and may prove to be a valid instrument in the detection of a wheelchair users’ physical activity.Implications for RehabilitationWhen placed on the hip of able-bodied persons, accelerometers have been shown to be a useful tool for assessing physical activity.Current literature of wheelchair users and accelerometry includes investigations of arm-mounted, body-mounted, and wheel-mounted accelerometers.The current study investigates a wheelchair-frame mounted triaxial accelerometer’s ability to detect group differences during light and moderate intensity wheeling under controlled conditions.Findings indicate that wheelchair frame-mounted accelerometers may be a useful tool for measuring a wheelchair user’s daily activity patterns and estimating energy expenditure in a free living environment.

Construct validity of RT3 accelerometer: A comparison of level-ground and treadmill walking at self-selected speeds

This study examined differences in accelerometer output when subjects walked on level ground and on a treadmill. We asked 25 nondisabled participants to wear an RT3 triaxial accelerometer (StayHealthy, Inc; Monrovia, California) and walk at their "normal" and "brisk" walking speeds for 10 minutes. These activities were repeated on a treadmill using the individual speeds from level-ground walking on two occasions 1 week apart. Paired t-tests found a difference in RT3 accelerometer vector magnitude (VM) counts/min between the two walking speeds on both surfaces on days 1 and 2 (p < 0.001). Although we found no significant differences between VM counts/min on the two surfaces at normal and brisk speeds on days 1 and 2 (p > 0.05), we found wide limits of agreement between level ground and treadmill walking at both speeds. Measurement and discrimination of walking intensity employing RT3 accelerometer VM counts/min on the treadmill demonstrated reasonable validity and stability over two time points compared with level-ground walking.