Epsilon-Tube Filtering: Reduction ofHigh-Amplitude Motion Artifacts FromImpedance Plethysmography Signal

Food Wastage Reduction Management Android App

Epsilon-Tube Filtering: Reduction of High-Amplitude Motion Artifacts From Impedance Plethysmography Signal

The impedance plethysmography (IP) has long beenused to monitor respiration. The IP signal is also suitable forportable monitoring of respiration due to its simplicity. However,this signal is very susceptible to motion artifact (MA). As a re-sult, MA reduction is an indispensable part of portable acquisitionof the IP signal. Often, the amplitude of the MA is much largerthan the amplitude of the respiratory component in the IP signal.This study proposes a novel filtering method to remove the high-amplitude MA’s from the IP signal. The proposed method combinesthe idea ofε-tube loss function and an autoregressive exogenousmodel to estimate the MA while leaving the periodic respiratorycomponent of the IP signal intact. Also, a regularization method isused to find the best filter coefficients that maximize the regularityof the output signal. The results indicate that the proposed methodcan effectively remove the MA, outperforming the popular MA re-duction methods. Several different performance measures are usedfor the comparison and the differences are found to be statisticallysignificant.
Motion artifact (MA) reduction is one of the most chal-lenging problems encountered during filtering and pro-cessing of physiological signals, especially those that are col-lected using portable monitoring devices. The main difficultyin dealing with MA is its dynamic nature and the fact that theamplitude of such artifacts is often much larger than the ampli-tude of the signal of interest. The previous studies that addressthis problem can be broadly divided into three groups. The firstgroup is focused on instrumentation.

https://codeshoppy.com/shop/product/a-food-wastage-reduction-mobile-application/

The aim of these studiesis to provide alternative instruments for signal recording, suchas electrodes, sensors, or wires, as well as different electrode placements that are less susceptible to MA [1], [2]. 
Such im-provements are outside the scope of this paper.The second group uses independent component analysis(ICA) and principal component analysis (PCA) to estimate theMA as an independent source of variation in the signal [3]–[6].The goal of ICA is to estimate the set of linear coefficientsthrough which the signal of interest and the MA have beenmixed. It requires multichannel recordings of the signal whichare often available for the electrocardiogram (ECG) and elec-troencephalography signals. However, the IP signal is typicallymeasured through a single channel recording. The multichannelrecording of the IP signal will require excessive instrumentationand hardware which is not desirable when designing a portabledevice. Therefore, ICA is not an appropriate candidate for MAreduction from the IP signal here. Another disadvantage of us-ing ICA is that the assumption of linear mixing process is notguaranteed to hold.The third group of MA reduction methods uses adaptive fil-ters to estimate the MA. Different varieties of adaptive filtershave been applied to this problem including least mean squares(LMS) [7]–[9], recursive least squares (RLS), [10], [11], nor-malized LMS (NLMS) [12], and normalized signed regressorLMS [13]. Despite the popularity of adaptive filters, there is adisadvantage associated with them. Adaptive filters tend to notonly model the MA, but they can also adapt to the signal of inter-est. In particular, the MA in the IP signal is similar in shape andfrequency bandwidth to the respiratory component of the signal.Therefore, an adaptive filter that successfully models the MAwill also model the respiratory component in the calm (not con-taminated by MA) sections of the signal. This is due to the factthat adaptive filters do not have a mechanism to distinguish be-tween the MA and the component of interest when they are sim-ilar. However, theε-tube filtering method that is proposed in thispaper focuses only on the MA and is not affected by the semis-inusoidal pattern of the respiratory component of the IP signal.The proposed filtering method is applied to the IP signal toeliminate the MA. Code Shoppy The main component of this signal is highlycorrelated with respiration which has a regular periodic patternwith an amplitude that is almost constant within a short windowof time. The amplitude of the interfering MA is often largerthan the amplitude of the respiratory component. Therefore,the proposed filtering method can be applied to this signal toremove the MA. The end-tidal CO2(EtCO2) signal is used as thereference signal in this study to compare the filter output to theactual respiration waveform. The EtCO2signal is considered thegold standard in monitoring the respiration. In order to show that theε-tube filtering, in fact, has a better performance comparedto the currently existing methods, it is compared to the ICAalgorithm, and the RLS and NLMS filtering methods in theresults section.