Font Size: a A A

The Key Technology Reasearch On Motion Sensors Measuring Phsical Activity

Posted on:2016-02-09Degree:DoctorType:Dissertation
Country:ChinaCandidate:J S DaiFull Text:PDF
GTID:1318330461490981Subject:Epidemiology and Health Statistics
Abstract/Summary:PDF Full Text Request
A large number of epidemiological studies have confirmed that physical activity plays a very important role in promoting health and reducing chronic diseases. The first step of launching physical activitity study is to develop objective, accurate and practical physical activity measurments in order to evaluate dose-response relationship between physical activity and health, to assess the validity about intervention of increasing physical activity and form public health gudieline based on objective evaluation. However, it is an challenge for researchers to measure physical activity accurately. Traditional methods, such as questionnaire survey, retrospective diary and so on, were almost not precise because those measures are difficult to avoid many subjective bias. So researchers are trying to develop objective measurements in order to appraise physical activity accurately. In many objective methods, motion sensor is the most widely used and most representative measurement on surveying physical activity. The aim of the study is to evaulate metabolic characteristics of Chinese typical physical activity, to assess the validity of five representative motion sensors(ActiGraph GT1 M Uniaxial Accelerometer, Acti Graph GT3 X Triaxial Accelerometer, RT3 Triaxial Accelerometer, Actiheart, IDEEA) on measuring Chinese typical physical activity, and to develop more precise forecast formulas which are more fit for measuring Chinese physical activity.Fourteen males and fifteen females completed 20 activities including 2mph walking, 3mph walking, 3mph 3% grade walking, 3mph 8% grade walking, 4mph brisk walking, 4mph 3% grade brisk walking, 5mph and 6mph running, slow and quick skipping, 10 mph and 13 mph bicycle riding, ascending stairs, descending stairs, cleaning table, washing clothes, mopping floor, playing table tennis, strength training for upper limb and lower limb. The volunteers worn ActiGraph GT1 M Uniaxial Accelerometer, ActiGraph GT3 X Triaxial Accelerometer, RT3 Triaxial Accelerometer, Actiheart and IDEEA according to standards. Oxygen consumption(VO2) was measured using a portable metabolic system named K4b2. MET values were defined as measured METs(VO2/measured resting metabolic rate) and standard METs(VO2/3.5 ml/kg/min).The first part of resluts showed Compendium METs were different than measured METs for 15/19 activities. Measured METs of most physical activities were higher than METs values from the compendium of physical activities. The number of activities different than the compendium was similar when standard METs were used(16/19). Oxygen uptake at rest was close to 3.5ml/kg/min among males so that measured METs were the same with standard METs, however, standard METs were lower than measured METs because oxygen uptake at rest was 3.0 ml/kg/min among females. No matter males or females, using standard METs may result in error classification of physical activity intensity and error classification rate is about 20%.the r2 between counts and METs was 0.79 when the speed of walking on the flat ground was below 5mph, otherwise, the r2 between counts and METs fell to 0.51 when considering all 20 activities. Since counts were possible lower or cann't increase along with activity intensity when measuring upper limb activity, walking on slope, going up and down the ladder, bicycle riding and so on, r2 between counts and METs declined when measuring these special types of activities. The most common equations such as Freedson equation, Swartz equation evidently underestimated energy expenditure of locomotion acticities and activities of daily living. Regression formulas according to walking and all activities in the research have more accuracy than the regression formulas of foreign, however, energy expenditure was overestimated in 8% of activities and underestimated in 43% of acitivities when regression formula according to walking was applied, on the otherwise, energy expenditure was overestimated in 20% of activities and underestimated in 23% of acitivities when regression formula according to all activities was applied.The r2 between counts of Vertical(V), medial–lateral(ML), anterior–posterior(AP) planes, vector magnitude and METs were 0.53?0.08?0.28,0.53. ML and AP planes of ActiGraph Triaxial Accelerometer couldn't accurately capture walking on slope besides vertical plane, however, the triaxial accelerometer could catch more physical activities based on the upper limb motion such as cleaning table, washing clothes, mopping floor, playing table tennis, strength training for upper limb compared with the uniaxial accelerometer. The mean errors of energy expenditure predictions were almost close between Freedson uniaxial regression equation and Freedson triaxial regression equation when examining 20 physical activities, the mean errors of energy expenditure prediction which was based on the Acti Graph triaxial accelerometer were lower than that of ActiGraph uniaxial accelerometer when measuring non-walk activities. Freedson triaxial regression equation evidently underestimated energy expenditure of locomotion acticities and activities of daily living. Regression formulas based on walking and all activities in the research have more accuracy than the regression formulas of foreign, however the equation based on the walking sometimes underestimated energy expenditure of walking and daily living activities, the equation based on all activities underestimated energy expenditure of walking and overestimated energy expenditure of daily living activities. In total, the equation based on all activities was more accurate than the equation based on the walking.The r2 between counts of vector magnitude, Vertical(V), medial–lateral(ML), anterior–posterior(AP) planes and METs were 0.63?0.62?0.59,0.46, the results showed three axis could occur synchronous changes with physical activities. The r2 between counts of vector magnitude counts, vertical counts and METs were almost same when measuring not only walking but also daily living activities. RT3 could sensitively measure flat walking, but couldn't accurately detect walking on slope, ascending and descending stairs. Since RT3 was worn besides the waist, it couldn't catch physical activities based on the upper limb motion such as cleaning table, washing clothes, mopping floor, playing table tennis, strength training for upper limb, however RT3 counld detect bicycle riding. RT3 regression equation evidently underestimated energy expenditure of locomotion acticities and activities of daily living. Regression formulas based on walking and all activities in the research have more accuracy than the regression formulas of the foreign, however the equation based on the walking sometimes underestimated energy expenditure of daily living activities, the equation based on all activities underestimated energy expenditure of moderate intensity activities. In total, the equation based on all activities was more accurate than the equation based on the walking.The original three regression models from the Actiheart software evidently underestimated energy expediture. The mean error of measuring walking was-1.65 which was the lowest from Actiheart activity algorithm, however, the mean errors of Actiheart HR algorithm and Actiheart combined activity and HR algorithm were-5.13 and-3.48 when meansuring walking. On the other hand, the mean errors of Actiheart HR algorithm and Actiheart combined activity and HR algorithm were-3.55 and-2.93 when meansuring daily living activities, which was lower than those of measuring walking. The means error increased to-2.17 when Actiheart activity algorithm was applied in meansuring daily living activities, which was higher than that of measuring walking. The mean errors from regression formulas which included HR algorithm, activity algorithm, combined activity and HR algorithm, individual HR algorithm from the research were lower than those of the original three formulas. The accuracy rate was beyond 70% when applying individual HR algorithm and combined activity and HR algorithm which were deduced by the research, but the accuracy rate was less than 20% when applying Actiheart HR algorithm and Actiheart combined activity and HR algorithm. The lowest mean error of individual HR algorithm was-0.09 when individual HR algorithm was applied in measuring walking. The highest accurate was from combined activity and HR algorithm when it was applied in measuring daily living activities.The IDEEA undetestimated METs of all activities and the mean error was 3METs. When measuring walking, the underestimation errors of METs forecast tended to increase along with walking speed and slope augment. On the other hand, IDEEA also underestimated METs of all daily living activities, furthermore, the errors of measuring moderate activities including cleaning table and washing clothes were lower than those of vigorous activities such as mopping floor, playing table tennis and strength training for lower limb.The study confirmed that regression formulas according to the research have more accuracy than the regression formulas of the foreign. The important study direction in the physical activity epidemiology is to develop motion sensors and infer core regression formulas which can measure accurately Chinese physical activity.
Keywords/Search Tags:physical activity, motion sensor, validity study
PDF Full Text Request
Related items