Algorithm to improve accuracy of energy expended in a room calorimeter

Abstract

The whole-room indirect calorimeter is considered as important equipment for human energy expenditure measurement, but noise reduction in the system remains a challenge. A selective filtering method (SFM) was designed to improve the accuracy of the computation of O₂ consumption rate ($\dot{V}_{{{\text{O}}_{ 2} }}$) and CO₂ production rate ($\dot{V}_{{{\text{CO}}_{ 2} }}$), based on two facts: (1) the rapid changes of $\dot{V}_{{{\text{O}}_{ 2} }}$, $\dot{V}_{{{\text{CO}}_{ 2} }}$ and respiratory quotient (RQ) in human should be accompanied by physical activity; (2) the oxygen consumption and the carbon dioxide production should not be negative because living humans do not generate oxygen, nor consume carbon dioxide. The performance of SFM was compared with the moving average method, the central difference method and the wavelet de-noising method. The range of $\dot{V}_{{{\text{O}}_{ 2} }}$ and $\dot{V}_{{{\text{CO}}_{ 2} }}$ in the empty room (the background noise) is reduced from −130.00–146.00 ml/min to −26.00–24.00 ml/min, and from −20.50–12.50 ml/min to −3.99–4.19 ml/min, by SFM. The background noise was added to simulated rectangular and sinusoidal signals that were used to evaluate the four methods over different time periods (64, 32, 16 and 8 min). The highest signal-to-noise ratio and the lowest deviation were achieved by SFM. Abnormal metabolic rates and RQs were corrected and compensated with measurement accuracy of 98.51 ± 0.3 % for 24-h alcohol burning tests. The results of the study showed that SFM can significantly improve $\dot{V}_{{\text{O}_{2} }}$ and $\dot{V}_{{{\text{CO}}_{ 2} }}$ measurements.

Glossary

RQ

Respiratory quotient

MR

Metabolic rate

SD

Standard deviation

$\dot{V}_{{{\text{O}}_{ 2} }}$

O₂ consumption rate

$\dot{V}_{{{\text{CO}}_{ 2} }}$

CO₂ production rate

$\dot{V}_{G}$

Input data set

$\dot{V}_{F}$

Filtering data set

SNR

Signal-to-noise ratio

SFM

Selective filtering method

CDM

Central difference method

MAM

Moving average method

WDM

Wavelet de-noising method