Mary Rosenberger

Activity recognition using a single accelerometer placed at the wrist or ankle.

PURPOSE Large physical activity surveillance projects such as the UK Biobank and NHANES are using wrist-worn accelerometer-based activity monitors that collect raw data. The goal is to increase wear time by asking subjects to wear the monitors on the wrist instead of the hip, and then to use information in the raw signal to improve activity type and intensity estimation. The purposes of this work was to obtain an algorithm to process wrist and ankle raw data and to classify behavior into four broad activity classes: ambulation, cycling, sedentary, and other activities. METHODS Participants (N = 33) wearing accelerometers on the wrist and ankle performed 26 daily activities. The accelerometer data were collected, cleaned, and preprocessed to extract features that characterize 2-, 4-, and 12.8-s data windows. Feature vectors encoding information about frequency and intensity of motion extracted from analysis of the raw signal were used with a support vector machine classifier to identify a subjects activity. Results were compared with categories classified by a human observer. Algorithms were validated using a leave-one-subject-out strategy. The computational complexity of each processing step was also evaluated. RESULTS With 12.8-s windows, the proposed strategy showed high classification accuracies for ankle data (95.0%) that decreased to 84.7% for wrist data. Shorter (4 s) windows only minimally decreased performances of the algorithm on the wrist to 84.2%. CONCLUSIONS A classification algorithm using 13 features shows good classification into the four classes given the complexity of the activities in the original data set. The algorithm is computationally efficient and could be implemented in real time on mobile devices with only 4-s latency.

Estimating activity and sedentary behavior from an accelerometer on the hip or wrist.

PURPOSE Previously, the National Health and Examination Survey measured physical activity with an accelerometer worn on the hip for 7 d but recently changed the location of the monitor to the wrist. This study compared estimates of physical activity intensity and type with an accelerometer on the hip versus the wrist. METHODS Healthy adults (n = 37) wore triaxial accelerometers (Wockets) on the hip and dominant wrist along with a portable metabolic unit to measure energy expenditure during 20 activities. Motion summary counts were created, and receiver operating characteristic (ROC) curves were then used to determine sedentary and activity intensity thresholds. Ambulatory activities were separated from other activities using the coefficient of variation of the counts. Mixed-model predictions were used to estimate activity intensity. RESULTS The ROC for determining sedentary behavior had greater sensitivity and specificity (71% and 96%) at the hip than at the wrist (53% and 76%), as did the ROC for moderate- to vigorous-intensity physical activity on the hip (70% and 83%) versus the wrist (30% and 69%). The ROC for the coefficient of variation associated with ambulation had a larger AUC at the hip compared to the wrist (0.83 and 0.74). The prediction model for activity energy expenditure resulted in an average difference of 0.55 ± 0.55 METs on the hip and 0.82 ± 0.93 METs on the wrist. CONCLUSIONS Methods frequently used for estimating activity energy expenditure and identifying activity intensity thresholds from an accelerometer on the hip generally do better than similar data from an accelerometer on the wrist. Accurately identifying sedentary behavior from a lack of wrist motion presents significant challenges.

Using wearable activity type detection to improve physical activity energy expenditure estimation

Accurate, real-time measurement of energy expended during everyday activities would enable development of novel health monitoring and wellness technologies. A technique using three miniature wearable accelerometers is presented that improves upon state-of-the-art energy expenditure (EE) estimation. On a dataset acquired from 24 subjects performing gym and household activities, we demonstrate how knowledge of activity type, which can be automatically inferred from the accelerometer data, can improve EE estimates by more than 15% when compared to the best estimates from other methods.

Twenty-four Hours of Sleep, Sedentary Behavior, and Physical Activity with Nine Wearable Devices.

UNLABELLED Getting enough sleep, exercising, and limiting sedentary activities can greatly contribute to disease prevention and overall health and longevity. Measuring the full 24-h activity cycle-sleep, sedentary behavior (SED), light-intensity physical activity (LPA), and moderate-to-vigorous physical activity (MVPA)-may now be feasible using small wearable devices. PURPOSE This study compared nine devices for accuracy in a 24-h activity measurement. METHODS Adults (n = 40, 47% male) wore nine devices for 24 h: ActiGraph GT3X+, activPAL, Fitbit One, GENEactiv, Jawbone Up, LUMOback, Nike Fuelband, Omron pedometer, and Z-Machine. Comparisons (with standards) were made for total sleep time (Z-machine), time spent in SED (activPAL), LPA (GT3X+), MVPA (GT3X+), and steps (Omron). Analysis included mean absolute percent error, equivalence testing, and Bland-Altman plots. RESULTS Error rates ranged from 8.1% to 16.9% for sleep, 9.5% to 65.8% for SED, 19.7% to 28.0% for LPA, 51.8% to 92% for MVPA, and 14.1% to 29.9% for steps. Equivalence testing indicated that only two comparisons were significantly equivalent to standards: the LUMOback for SED and the GT3X+ for sleep. Bland-Altman plots indicated GT3X+ had the closest measurement for sleep, LUMOback for SED, GENEactiv for LPA, Fitbit for MVPA, and GT3X+ for steps. CONCLUSIONS Currently, no device accurately captures activity data across the entire 24-h day, but the future of activity measurement should aim for accurate 24-h measurement as a goal. Researchers should continue to select measurement devices on the basis of their primary outcomes of interest.

Feasibility of Obtaining Measures of Lifestyle From a Smartphone App: The MyHeart Counts Cardiovascular Health Study

Importance Studies have established the importance of physical activity and fitness, yet limited data exist on the associations between objective, real-world physical activity patterns, fitness, sleep, and cardiovascular health. Objectives To assess the feasibility of obtaining measures of physical activity, fitness, and sleep from smartphones and to gain insights into activity patterns associated with life satisfaction and self-reported disease. Design, Setting, and Participants The MyHeart Counts smartphone app was made available in March 2015, and prospective participants downloaded the free app between March and October 2015. In this smartphone-based study of cardiovascular health, participants recorded physical activity, filled out health questionnaires, and completed a 6-minute walk test. The app was available to download within the United States. Main Outcomes and Measures The feasibility of consent and data collection entirely on a smartphone, the use of machine learning to cluster participants, and the associations between activity patterns, life satisfaction, and self-reported disease. Results From the launch to the time of the data freeze for this study (March to October 2015), the number of individuals (self-selected) who consented to participate was 48 968, representing all 50 states and the District of Columbia. Their median age was 36 years (interquartile range, 27-50 years), and 82.2% (30 338 male, 6556 female, 10 other, and 3115 unknown) were male. In total, 40 017 (81.7% of those who consented) uploaded data. Among those who consented, 20 345 individuals (41.5%) completed 4 of the 7 days of motion data collection, and 4552 individuals (9.3%) completed all 7 days. Among those who consented, 40 017 (81.7%) filled out some portion of the questionnaires, and 4990 (10.2%) completed the 6-minute walk test, made available only at the end of 7 days. The Heart Age Questionnaire, also available after 7 days, required entering lipid values and age 40 to 79 years (among 17 245 individuals, 43.1% of participants). Consequently, 1334 (2.7%) of those who consented completed all fields needed to compute heart age and a 10-year risk score. Physical activity was detected for a mean (SD) of 14.5% (8.0%) of individuals’ total recorded time. Physical activity patterns were identified by cluster analysis. A pattern of lower overall activity but more frequent transitions between active and inactive states was associated with equivalent self-reported cardiovascular disease as a pattern of higher overall activity with fewer transitions. Individuals’ perception of their activity and risk bore little relation to sensor-estimated activity or calculated cardiovascular risk. Conclusions and Relevance A smartphone-based study of cardiovascular health is feasible, and improvements in participant diversity and engagement will maximize yield from consented participants. Large-scale, real-world assessment of physical activity, fitness, and sleep using mobile devices may be a useful addition to future population health studies.

Validation of Physical Activity Tracking via Android Smartphones Compared to ActiGraph Accelerometer: Laboratory-Based and Free-Living Validation Studies.

Background There is increasing interest in using smartphones as stand-alone physical activity monitors via their built-in accelerometers, but there is presently limited data on the validity of this approach. Objective The purpose of this work was to determine the validity and reliability of 3 Android smartphones for measuring physical activity among midlife and older adults. Methods A laboratory (study 1) and a free-living (study 2) protocol were conducted. In study 1, individuals engaged in prescribed activities including sedentary (eg, sitting), light (sweeping), moderate (eg, walking 3 mph on a treadmill), and vigorous (eg, jogging 5 mph on a treadmill) activity over a 2-hour period wearing both an ActiGraph and 3 Android smartphones (ie, HTC MyTouch, Google Nexus One, and Motorola Cliq). In the free-living study, individuals engaged in usual daily activities over 7 days while wearing an Android smartphone (Google Nexus One) and an ActiGraph. Results Study 1 included 15 participants (age: mean 55.5, SD 6.6 years; women: 56%, 8/15). Correlations between the ActiGraph and the 3 phones were strong to very strong (ρ=.77-.82). Further, after excluding bicycling and standing, cut-point derived classifications of activities yielded a high percentage of activities classified correctly according to intensity level (eg, 78%-91% by phone) that were similar to the ActiGraph’s percent correctly classified (ie, 91%). Study 2 included 23 participants (age: mean 57.0, SD 6.4 years; women: 74%, 17/23). Within the free-living context, results suggested a moderate correlation (ie, ρ=.59, P<.001) between the raw ActiGraph counts/minute and the phone’s raw counts/minute and a strong correlation on minutes of moderate-to-vigorous physical activity (MVPA; ie, ρ=.67, P<.001). Results from Bland-Altman plots suggested close mean absolute estimates of sedentary (mean difference=–26 min/day of sedentary behavior) and MVPA (mean difference=–1.3 min/day of MVPA) although there was large variation. Conclusions Overall, results suggest that an Android smartphone can provide comparable estimates of physical activity to an ActiGraph in both a laboratory-based and free-living context for estimating sedentary and MVPA and that different Android smartphones may reliably confer similar estimates.

Activity Recognition in Youth Using Single Accelerometer Placed at Wrist or Ankle.

Purpose State-of-the-art methods for recognizing human activity using raw data from body-worn accelerometers have primarily been validated with data collected from adults. This study applies a previously available method for activity classification using wrist or ankle accelerometer to data sets collected from both adults and youth. Methods An algorithm for detecting activity from wrist-worn accelerometers, originally developed using data from 33 adults, is tested on a data set of 20 youth (age, 13 ± 1.3 yr). The algorithm is also extended by adding new features required to improve performance on the youth data set. Subsequent tests on both the adult and youth data were performed using crossed tests (training on one group and testing on the other) and leave-one-subject-out cross-validation. Results The new feature set improved overall recognition using wrist data by 2.3% for adults and 5.1% for youth. Leave-one-subject-out cross-validation accuracy performance was 87.0% (wrist) and 94.8% (ankle) for adults, and 91.0% (wrist) and 92.4% (ankle) for youth. Merging the two data sets, overall accuracy was 88.5% (wrist) and 91.6% (ankle). Conclusions Previously available methodological approaches for activity classification in adults can be extended to youth data. Including youth data in the training phase and using features designed to capture information on the activity fragmentation of young participants allows a better fit of the methodological framework to the characteristics of activity in youth, improving its overall performance. The proposed algorithm differentiates ambulation from sedentary activities that involve gesturing in wrist data, such as that being collected in large surveillance studies.