Effects of Frequency Filtering on Intensity and Noise in Accelerometer-Based Physical Activity Measurements

Abstract

In objective physical activity (PA) measurements, applying wider frequency filters than the most commonly used ActiGraph (AG) filter may be beneficial when processing accelerometry data. However, the vulnerability of wider filters to noise has not been investigated previously. This study explored the effect of wider frequency filters on measurements of PA, sedentary behavior (SED), and capturing of noise. Apart from the standard AG band-pass filter (0.29–1.63 Hz), modified filters with low-pass component cutoffs at 4 Hz, 10 Hz, or removed were analyzed. Calibrations against energy expenditure were performed with lab data from children and adults to generate filter-specific intensity cut-points. Free-living accelerometer data from children and adults were processed using the different filters and intensity cut-points. There was a contribution of acceleration related to PA at frequencies up to 10 Hz. The contribution was more pronounced at moderate and vigorous PA levels, although additional acceleration also occurred at SED. The classification discrepancy between AG and the wider filters was small at SED (1–2%) but very large at the highest intensities (>90%). The present study suggests an optimal low-pass frequency filter with a cutoff at 10 Hz to include all acceleration relevant to PA with minimal effect of noise.

Stronger Association between High Intensity Physical Activity and Cardiometabolic Health with Improved Assessment of the Full Intensity Range Using Accelerometry

“…The main result of the present study was that when applying the new method with the 10 Hz filter, which is wider than the most commonly used AG filter, to the processing of raw acceleration data the association between PA and cardiometabolic health was shifted towards higher intensities. Although both models show a peak of the relationship at an intensity higher than regular walking, the peak of the AG output was equivalent to the walking-running transition whereas the peak of the 10 Hz output was at higher running speeds (Figures 2 and 3) [7]. More importantly, the association did not decline immediately with higher running speeds with the 10 Hz output as with the AG output.…”

“…Lab based results show that there is no association between high AG output and energy expenditure whereas with the 10 Hz filter the association remains at high intensity [6]. Previous results also suggests that AG heavily overestimates the amount of high intensity PA compared to 10 Hz [7]. Since lab based results suggests that the 10 Hz filter better captures high intensity PA, high intensity 10 Hz output should be considered of higher quality and closer to the true PA than the AG output at the same intensity.…”

“…Extracted raw acceleration was processed to the output mean mg of 3 s epochs using either the standard AG filter or a modified filter with a 10 Hz low-pass cut-off. Technical details of the processing has been published previously [7]. Night time between 00:00 and 06:00 was removed from the analysis.…”

“…A valid measurement was considered at least four valid days, which in turn was defined as at least eight hours of wear-time [9]. Previously published energy expenditure cut-points for 1.5, 3, 6 and 9 METs representing LPA, MPA, VPA and VVPA intensity was considered reference for PA intensity [7]. The AG cut-points were 19.1, 63.0, 171.0 and 300.2 mg and the 10 Hz cut-points were 38.9, 167.2, 582.3 and 994.1 mg. To allow for more detailed analyses of the PA intensity spectrum, the intensity spectrum was divided into smaller bins.…”

“…With this comparison it is also apparent that, relative to the 10 Hz filter, the AG output yields much more high intensity PA. This can be explained by the lab results showing that the relationship between AG output and energy expenditure is very weak at high intensity, suggesting that the epochs are mainly classified by random [7].…”