Alcohol abuse is the third leading lifestyle-related cause of death for individuals in the United States, causing 88,000 deaths each year in the United States from 2006-2010. Existing smartphone applications allow users to manually record their alcohol consumption or take cognitive tests to estimate intoxication levels; however, no smartphone application passively determines one's level of intoxication. After gathering smartphone sensor data from 34 "intoxicated" subjects, we generated time and frequency domain features such as sway area (gyroscope) and cadence (accelerometer), which were then classified using a supervised machine learning framework. Other novel contributions explored include feature normalization to account for differences in walking styles and automatic outlier elimination to reduce the effect of accidental falls by identifying and removing the top and bottom of a chosen percentage of the data. Various machine learning classifier types such as Random Forest and Bayes Net were compared, and J48 classifier was the most accurate, classifying user gait patterns into BAC ranges of [0.00-0.08), [0.08-0.15), [0.15-0.25), [0.25+) with an accuracy of 89.45%. This best performing classifier was used to build an intelligent smartphone app that will detect the user's intoxication level in real time.

Itch lies second only to disturbance of body image as a reported symptom in dermatology. This study started by concentrating on improving the measurement of itch. Itch has a paired physical response, scratch. The pairing can be exploited: preliminary work by this unit had validated the use of wrist-worn movement-measuring machines called ‘accelerometers’ to measure itch-related movement (scratch and rub). The first part of this research developed use of these machines. Simple accelerometers (‘Actiwatch Plus’) were used to observe the pattern of variation of itch over clusters of nights and in different conditions. The accelerometer scores were able to identify controls’ scores from those with itchy disease. Considerable variation (56%) was discovered in objective score between subject and considerable variation was noted (46%) even within subject. More complex accelerometers, (‘DigiTrac’) which could potentially specifically identify itch-related movement on the basis of frequency of action derived from Fast Fourier Transform (FFT), were validated against the ‘gold standard’ measurement of itch-related movement, directly observed movement (via infra red video recording). It was necessary to characterise the ‘frequency of action’ of itch on video and, as an aside, the characteristics of human itch-related movement were compared to other mammals’ itch-related movement ‘frequency of action’. The ‘frequency of action’ and video data was used to enrich the DigiTrac readouts to improve specificity of itch-related movement detection. During the accelerometer studies, an unexpected finding came to light: objective score of itch was not related to subjective score. To try to explain the lack of relationship, a 42 day longitudinal study of atopic dermatitis patients’ subjective and objective scores was undertaken. The results demonstrated autocorrelation for subjective scores, but not for the objective scores but still did not fully explain the lack of relationship. In an effort to explain the disconnect between subjective and objective scores a second tranche of experiments and the second part of this research interrogated whether the methods with which we measure disease as a whole in dermatology are robust. One study investigated whether the way patients are asked about subjective symptoms in general was resistant to the effects of focusing and framing bias. The results were reassuring as they suggested that the commonly used and recommended symptom scoring systems were robust in the face of bias. In order to assess whether perspective or perception of disease explained the disconnect, a study was designed in collaboration with the Edinburgh College of Art. A series of computer-generated images of different psoriasis severities were created and used to assess how doctors and patients assessed disease-extent. This study showed that, whilst each group had a naturally divergent opinion of extent of disease, by scoring disease using the models it was possible to unify the perspective and perception of extent. Finally, an exploratory study to reduce recall bias to a minimum, in case this had caused the disconnect between objective and subjective, was undertaken. This employed a novel questionnaire, the Day Reconstruction Method.

PURPOSE: To understand the influences of mechanical loading on bone adaptation, the ground reaction force (GRF) applied to the bone must be quantified. The use of force plates in a lab setting is the accepted method for quantifying GRFs; however, this is not feasible in free-living situations. Recent developments in accelerometer technology may provide the ability to evaluate the effects of mechanical loading on bone outside of laboratory settings. The purpose of this project was to validate an accelerometer for the measurement of mechanical loading by comparing its output against GRFs. METHODS: Male and female participants (n = 20 males, 20 females; 18 to 49 yr) completed 10 repetitions of 9 common everyday movements (stand, walk, jog, run, 15 cm jump, step down from curb, drop down from curb, forward hop, and side hop) on a force plate with an accelerometer worn on their right hip. Then, a subset (n = 5 males, 5 females) wore an accelerometer on their right hip and played basketball, volleyball, and dodgeball as a group. Finally, all 40 participants wore an accelerometer home for 7 days. All activities were organized into derived activity categories labeled as low-, moderate-, and high-mechanical-load-intensity and used with 59 possible accelerometer variables to predict mechanical load. Models were fit using the randomForest package in R. Model performance (coefficient of determination [R2] and median absolute error) was evaluated using cross-validation. RESULTS: The percentage of variation mechanical load intensity explained by the models ranged from 0.27 to 0.78 with median absolute errors ranging from 0.20 to 0.49. The model with R2 = 0.78 contained the known activity categories and the accelerometer variables, but this is not realistic for free-living situations where activity categories will not be known. The two free-living models with the highest R2 values included derived activity categories and accelerometer variables, and estimated, on average, 21.1 and 20.7 hours per day in low-intensity, 1.6 and 1.7 hours per day in moderate-intensity, and 0.0 and 0.5 hours per day in high-intensity osteogenic activity, respectively. CONCLUSION: It is assumed that higher intensity activities (i.e., jumping vs. jogging) result in higher GRF values, but depending on the actual execution of the movement, this is not always the case. This research demonstrated that models containing the accelerometer variables performed better in predicting GRF than those containing only the derived activity categories. This supports the hypothesis that accelerometers provide valuable objective information when evaluating mechanical loading on bone.

80% of the American population experiences back pain and it is the most common cause of limited activity in people of age 45 and under. Determining the reasons for back pain and developing new ways to treat it have been extensively researched over the past decade. However, very little research has been done on low back pain of amputees. There are four million existing amputees living in America and 250,000 people become new amputees each year. 70% of this group is lower limb amputees and a large number use a prosthesis of some kind to aid in the functions of daily living (Amputation and Limb Deficiency). Not all amputees use a prosthesis because of pain involved, aesthetics, and cost. In order to increase the use of prosthetics among amputees, the reasons why they do not use them must be fully understood. With this knowledge better prosthetic designs can be created. The purpose of this study is to first determine the prevalence of back pain among lower limb amputees who use a prosthesis and then to quantify the accelerations in the spine of this group and compare it to subjects who are not amputees. The findings of this study will be used to determine if back pain is a common complaint, if it interferes with daily activities, and if the use of a prosthesis causes abnormal loads in the spine of amputees. A cross-sectional descriptive survey was created and distributed to lower limb amputees who use a prosthesis and to a control group. In addition to the survey, several subjects were recruited to wear an accelerometer located over the L5-S1 vertebrae and walk at several speeds down a pathway. A maximum acceleration was determined for each step as well as the difference in acceleration between opposing legs. Also measured was the effect of a leg length discrepancy (LLD) on accelerations and back pain. As a result of this research it was found that a high percentage of amputees experience back pain and the prevalence is higher than that of controls. It has shown that there is a difference between the acceleration patterns of amputees and non-amputees, but further research is needed to show that this difference is what causes the higher prevalence of back pain. The trend of side dominance and its increase with increased walking speed for amputees has been shown as well as a general population trend of increased acceleration of the spine with increased speed. In relation to walking speed, the study has also shown that the perception of speed among amputees is slower than that of controls. This study has also supported the notion that a difference in leg length could cause low back pain.

The aim of this work was to develop devices for elite athletes to record performance related parameters during their training. A device was initially designed and built for rowing to record the motion of the boat. This was to gain understanding of motion signals in a one dimensional plane. The device uses a iPAQ handheld computer for recording and display of data to the user. Using the knowledge obtained from the accelerometer data of the rowing system an initial prototype device was designed and constructed for use in swimming. This device was required to be wearable whilst the swimmer was training, thus it had to record the data onboard. A second version of the swimming device was constructed to improve the usability of the device. The swimming device has fully sealed electronics, wireless charging and infrared communications. The device records three dimensional acceleration patterns at 150Hz, and can store over 6 hours of data using the internal memory. The device can operate for greater than 12 hours before needing to be recharged. The data collected from the swimming device was used to develop processing algorithms to extract when the swimmers push off from the wall, the type of stroke they are swimming, and for freestyle the stroke count. The results of the wall push off algorithm were compared against manual hand timing with 90% algorithm results being with ±1 second of the hand timing data. The stroke type identification algorithm determines which stroke is being swum and presently has an accuracy of 95%. The results of the freestyle stroke count algorithm were compared against manual stroke counts from raw accelerometers data and underwater video. Of the 164 data sets analysed over 90% of the algorithm results were within ±1 strokes of the manual recorded stroke counts.

<p>The Automated Manual Transmission system Opticruise is dependent on signals from sensors located in different parts of the Scania trucks. These signals are of different qualities and have different update frequencies. Some signals and quantities that are hard or impossible to measure are also of importance to this system.</p><p>In this thesis a driveline observer for the purpose of signal improvement is developed and estimations of unknown quantities such as road incline and mass of the vehicle are performed. The outputs of the observer are produced at a rate of 100 Hz, and include in addition to the mass and road incline also the speed of the engine, output shaft of the gearbox, wheel and the torsion in the driveline. Further the use of an accelerometer and the advantages gained from using it in the observer are investigated.</p><p>The outputs show an increased quality and much of the measurement noise is successfully removed without introducing any time delays. A simulation frequency of 100 Hz is possible, but some dependency toward the stiffness of the driveline is found. The observer manages to estimate the road slope accurately. With the use of an accelerometer the road slope estimation is further improved and a quickly converging mass estimation is obtained.</p>

Purpose: To determine the step count accuracy of three pedometers and one accelerometer in pregnant women during treadmill walking. Methods: Subjects were 30 women in the second or third trimester (20-36 weeks) who were screened for pregnancy-related risk factors. Each subject was fitted with a belt containing three physical activity monitors: Yamax Digiwalker SW-200 (DW), New Lifestyles NL 2000 (NL), and GT3X Actigraph accelerometer (ACT). The Omron HJ-720 (HJ) was placed in the pants pocket. Subjects walked at 54, 67, 80, and 94 m•min-1 for two minutes each. Actual steps were determined by an investigator using a hand-tally counter. Percentage of actual steps was calculated for each device at each speed and compared. Results: There was a significant interaction between speed and device (F9,20=7.574,P<0.001). At all speeds, the NL and HJ were most accurate. At 54 m•min-1, the DW was significantly less accurate (P<0.001) than all other devices and the ACT was significantly less accurate (P<0.001) than the NL and HJ. At 67 m•min-1, the ACT and DW were significantly less accurate (P<0.001) than the NL and HJ. At 80 m•min-1, the DW was significantly less accurate (P=0.024) than the NL and HJ. At 94 m•min-1, the ACT was significantly less accurate (P=0.001) than the NL and HJ. No significant differences were found at any speed for the NL (P=0.996) and HJ (P=0.298). Trimester did not significantly affect device accuracy. Conclusion: In pregnant women, the ACT and DW are less accurate than the NL and HJ. The HJ appeared to be the most accurate. These results can be useful in developing further research studies and physical activity programs that focus on walking during pregnancy.

Advances in electronic sensor technologies have led to the increased use of accelerometers for measuring physical activity and sedentary behaviours. Accelerometers overcome many of the inherent limitations of other measurement methods; for example, unlike self-reported instruments, accelerometers are free from random and systematic errors introduced by respondents and interviewers, cultural tradition, and language. However, accelerometers have their own set of limitations; for example, not all accelerometers are created equal and raw accelerometer data require significant data mining procedures in order to yield meaningful outcome variables. Therefore the overall purpose of this three study dissertation was to determine the impact accelerometer model has on the development of a comprehensive physical activity and sedentary behaviour profile and to design and apply novel profiling methods in an order to gain new insights into childrens physical activity.<p> Study One Purpose: To determine which of the three most commonly used accelerometer models has the best intra- and inter-instrument reliability using a mechanical laboratory setup. Secondly, to determine the effect acceleration and frequency have on these reliability measures. Methods: Three experiments were performed. In the first, five each of the Actical, Actigraph, and RT3 accelerometers were placed on a hydraulic shaker plate and simultaneously accelerated in the vertical plane at varying accelerations and frequencies. Six different conditions of varying intensity were used to produce a range of accelerometer counts. Reliability was calculated using standard deviation, standard error of the measurement, coefficient of variation, and intraclass correlation coefficients. In the second and third experiment, 39 Actical and 50 Actigraph accelerometers were put through the same six conditions. Results: Experiment One showed poor reliability in the RT3 (intra- and inter-instrument CV > 40%). Experiments Two and Three clearly indicated that the Actical (CVintra = 0.5%; CVinter = 5.4%) was more reliable than the Actigraph (CVintra = 3.2%; CVinter = 8.6%). Variability in the Actical was negatively related to the acceleration of the condition while no relationship was found between acceleration and reliability in the Actigraph. Variability in the Actigraph was negatively related to the frequency of the condition while no relationship was found between frequency and reliability in the Actical. Conclusion: Of the three accelerometer models measured in this study, the Actical had the best intra- and inter-instrument reliability. However, discrepant trends in the variability of Actical and Actigraph counts across accelerations and frequencies preclude the selection of a superior model. More work is needed to understand why accelerometers designed to measure the same thing, behave so differently.<p> Study Two The accurate measurement of habitual physical activity is fundamental to the study of the relationship between physical activity and health. However, many physical activity measurement techniques produce variables accurate to only the day level, such as total energy expenditure via self-report questionnaire, pedometer step counts or accelerometer measurements of minutes of moderate to vigorous physical activity. Monitoring technologies providing more detailed information on physical activity/sedentary behaviour can now be used to explore the relationships between health and movement frequency, intensity, and duration more comprehensively. This paper explores the activity and sedentary profile that can be acquired through objective monitoring, with a focus on accelerometry. Using previously collected objective data, a detailed physical activity profile is presented and case study examples of data utilization and interpretation are provided. The rich detail captured through comprehensive profiling creates new surveillance and study possibilities and could inform new physical activity guidelines. Data are presented in various formats to demonstrate the dangers of misinterpretation when monitoring population adherence to Canadas Physical Activity Guidelines. Recommendations for physical activity and sedentary profiling are provided and future research needs identified.<p> Study Three Purpose: This study explored the influence of modernity on the physical activity behaviours (e.g. intensity and timing) of children. Methods: Children aged 8-13 years living a traditional lifestyle (Old Order Amish; OOA n=68, Old Order Mennonite; OOM n=120) were compared with children living a contemporary lifestyle (rural Saskatchewan; RSK n=132 and urban Saskatchewan; USK n=93). Physical activity was objectively assessed for seven consecutive days using Actigraph 7164 accelerometers. Custom software was used to reduce the raw accelerometer data into standardized outcome variables. Results: On weekdays there were group differences in moderate physical activity between all lifestyle groups (OOA > OOM > USK > RSK). On the weekend, the group differences in moderate physical activity persisted between, but not within, lifestyle groups (OOA = OOM > USK = RSK). During school hours, all groups had similar activity and sedentary timings; however, they differed in magnitude with the OOA and OOM being both more sedentary and more active. Compared to in school, the OOA and OOM children had 44% lower sedentary time out of school compared to only 15% lower for RSK and USK children. Conclusions: Though cross-sectional, these data suggest that contemporary/modern living is associated with lower levels of moderate and vigorous intensity physical activity compared to lifestyles representative of earlier generations. Analyzing the physical activity and sedentary patterns of traditional lifestyle groups such as the OOA and OOM can provide valuable insight into the quantity and quality of physical activity necessary to promote health.<p> General Conclusions: Together, these three studies will help contribute to the generation of best practices in the accelerometric profiling of both physical activity and sedentary behaviours.

The metabolic syndrome has become a major public health challenge world-wide and, at least in the industrialized world, the prevalence of the metabolic syndrome is increasing. There is evidence to show that biological and lifestyle risk factors for metabolic syndrome are present in adolescence, which suggests that the antecedents of the disease may lie in early life. The period of adolescence is characterized by a decline in physical activity (PA; lack of PA is a lifestyle risk factor for metabolic syndrome) and an increase in fat mass deposition (a biological risk factor for metabolic syndrome). Therefore, investigating how the development of these two variables relates to adult cardiometabolic risk is important to fuel early intervention. A factor which has the potential to influence these two risk factors, and thus ultimately the metabolic syndrome, is the timing of biological maturity (i.e. whether an individual is early, average or late maturing when compared to peers of the same age). The influence of biological maturity has largely been overlooked in previous research; therefore, the general objective of this thesis was to investigate the associations between biological maturity, adolescent PA and fat mass development, and young adult cardiometabolic risk. Three studies were necessary to realize this objective, and together help to elucidate the role of biological maturity in the adolescent decline in physical activity, fat development, and the development of adult metabolic syndrome. Ultimately, this information will aid in the development and implementation of interventions to decrease prevalence of metabolic syndrome.<p> Study 1: The purpose of study 1 was to investigate whether observed gender differences in objectively measured PA in children (8 to 13 years) are confounded by biological maturity differences. Methods: Four hundred and one children (194 boys and 207 girls) volunteered for this study. An Actigraph accelerometer was used to obtain 7 consecutive days of minute-by-minute PA data on each participant. Minutes of moderate to vigorous PA per day (MVPA), continuous minutes of MVPA per day (CMVPA), and minutes of vigorous PA per day (VPA) were derived from the accelerometer data. Age at peak height velocity (APHV), an indicator of somatic maturity, was predicted and individuals aligned by this biological age (years from APHV). Gender differences in the PA variables were analyzed using a two-way (gender X age) ANOVA. Results: Levels of PA decreased with increasing chronological ages in both genders (p<0.05). When aligned on chronological age, boys had a higher MVPA at 10 through 13 years, a higher CMVPA at 9 through 12 years, and a higher VPA at 9 though 13 years (p<0.05). When aligned on biological age, PA declined with increasing maturity (p<0.05); however gender differences between biological age groups disappeared. Conclusion: The observed age-related decline in adolescent boys and girls PA is antithetical to public health goals and as such is an important area of research. In order to fully understand gender disparities in PA, consideration must be given to the confounding effects of biological maturity.<p> Study 2: Understanding the influence of biological age (BA) on the decline in PA would better inform researchers about the effective timing of intervention. The purpose of study 2 was to describe the PA levels and perceived barriers to PA of adolescent girls grouped by school grade and biological maturity status (i.e., early or late maturing) within grades. Methods: 221 girls (aged 8-16 years; grades 4-10) wore an Actical accelerometer for 7 days and then completed a semi-structured, open ended questionnaire on perceived barriers to PA over the 7 day period. Predicted APHV and recalled age at menarche were used to assess maturity among the elementary and high school girls, respectively. Maturity and grade group differences in PA were assessed using MANCOVA and independent sample t-test, and barriers to PA using chi squared statistics. Results: Daily minutes spent in MVPA decreased by 40% between grades 4 to 10. Within grade groupings, no differences in PA were found between early and late maturing girls (p>0.05). Grades 4-6 participants cited more interpersonal (i.e., social) barriers. Grades 9-10 participants cited more institutional barriers to PA, primarily revolving around the institution of school. No differences were found in types of barriers reported between early and late maturing girls. Conclusion: Since PA and types of perceived barriers to PA were dependent on grade, future research should work to identify the most salient (i.e., frequent and limiting) barriers to PA by chronological age in youth.<p> Study 3: Although the metabolic syndrome is thought to be mainly a consequence of obesity, the mechanisms underpinning its development are not that well understood. The purpose of study 3 was to examine total body fat mass (FM), trunk FM and PA developmental trajectories (aligned to BA; years from APHV) of individuals categorized as low and high for cardiometabolic risk at 26 years, while investigating biological and lifestyle risk factors. Methods: The sample were 55 males and 76 females from the Saskatchewan Pediatric Bone Mineral Accrual Study (1991-2007) who were assessed from childhood to young adulthood and had a measure of cardiometabolic risk at young adulthood (26.0 + 2.3 yrs). Height was measured biannually. Total body FM and trunk FM was assessed annually by dual energy-X-ray absorptiometry. PA and dietary intake was evaluated two to three times annually using surveys. Individuals were grouped into maturity status groups (early, average or late) depending on their APHV. Two composite cardiometabolic risk scores were calculated for males and females separately. The first was derived for a sub-sample (N=48) by summing the standardized residuals of inverted high-density lipoprotein cholesterol, homeostasis model assessment for insulin resistance, mean arterial pressure (MAP) and fasting triglyceride levels. A second score was derived for the whole sample by summing the standardized residuals for MAP. Scores for both samples were regressed on to age and adult smoking status. High and low cardiometabolic risk groups were determined based on a sex- specific median split of risk scores. Data were analyzed using random effects models. Models were built in a stepwise procedure with predictor variables added one at a time, using the log likelihood ratio statistic to determine if one model was a significant improvement over the previous one. Results: The final model indicated that once the independent effects of maturity (years from APHV) and height were controlled, the high risk group males and females had significantly (p<0.05) greater total body FM and trunk FM development at all ages. No association was found between young adult cardiometabolic risk and development of PA. Furthermore, in general, timing of biological maturity was not associated with development of PA or FM. Conclusion: Young adults at higher cardiometabolic risk have greater body fat as early as 8 years of age, which lends support to early intervention.<p> General Conclusions: Adolescence has been highlighted as a critical period for the development of adult disease, such as the metabolic syndrome. Results from this thesis support this contention by showing a decrease in PA (by both chronological and biological age) in males and females across adolescence. It further showed that an increase in total and central fatness during adolescence may be critical for the development of the metabolic syndrome in adulthood. Timing of biological maturity, in general, was not shown to have an independent impact on adolescent or young adult PA, adolescent perceived barriers to PA, fat mass development, or young adult cardiometabolic risk. However, further research is required before definitive conclusions can be made about the short and long term impacts of timing of biological maturity on health.

In the automatic manual transmission system, Opticruise, the choice of gear is based on several parameters such as road incline, driving resistance and vehicle mass. Many different mass estimations are made during driving. A final vehicle mass is then used to determine the current gear. Construction vehicles are often not equipped with air suspension and can therefore not estimate the vehicle mass when standing still. If that sort of vehicle is reloaded while standing still an incorrect mass estimation will be used and as an effect of that also a wrong gear. The solution of this problem is divided into two parts: first to detect the reload and then to estimate the new mass. An accelerometer will be used to detect a reload before start, because it reacts on the gradient change of the vehicle. A force equation will also be used to detect a reload, but only after the start, because it needs access to the vehicle acceleration to make the calculation. After a reload has been detected a new mass can be estimated. The estimation is based on previous vehicle mass estimated during driving. The new mass will be reasonable, if the assumption that the vehicle is mostly driving empty or fully loaded is true, and if it is possible to determine whether the vehicle becomes lighter or heavier at the reload.