Reduction In Mean Absolute Error Research Articles

Pedotransfer Function for the Brunswick Soil Hydraulic Property Model and Comparison to the van Genuchten‐Mualem Model

AbstractModeling soil hydraulic properties requires an effective representation of capillary and noncapillary storage and conductivity. This is made possible by using physically comprehensive yet flexible soil hydraulic property models. Such a model (Brunswick [BW] model) was introduced by Weber et al. (2019, https://doi.org/10.1029/2018WR024584), and it overcomes some core deficiencies present in the widely used van Genuchten‐Mualem (VGM) model. We first compared the performance of the BW model to that of the VGM model in its ability to describe water retention and hydraulic conductivity data on a set of measurements of 402 soil samples with textures covering the entire range of classes. Second, we developed a simple transfer function to predict BW parameters based on VGM parameters. Combined with our new function, any existing pedotransfer function for the prediction of the VGM parameters can be extended to predict BW model parameters. Based on information criteria, the smaller variance of the residuals, and a 40% reduction in mean absolute error in the hydraulic conductivity over all samples, the BW model clearly outperforms VGM. This is possible as the BW model explicitly accounts for hydraulic properties of dry soils. With the new pedotransfer function developed in this study, better descriptions of water retention and hydraulic conductivities are possible. We are convinced that this will strengthen the utility of the new model and enable improved field‐scale simulations, climate change impact assessments on water, energy and nutrient fluxes, as well as crop productivity in agroecosystems by soil‐crop and land‐surface modeling. The models and the pedotransfer function are included in an R package spsh (https://cran.r‐project.org/package=spsh).

Domain Agnostic Real-Valued Specificity Prediction

Sentence specificity quantifies the level of detail in a sentence, characterizing the organization of information in discourse. While this information is useful for many downstream applications, specificity prediction systems predict very coarse labels (binary or ternary) and are trained on and tailored toward specific domains (e.g., news). The goal of this work is to generalize specificity prediction to domains where no labeled data is available and output more nuanced realvalued specificity ratings.We present an unsupervised domain adaptation system for sentence specificity prediction, specifically designed to output real-valued estimates from binary training labels. To calibrate the values of these predictions appropriately, we regularize the posterior distribution of the labels towards a reference distribution. We show that our framework generalizes well to three different domains with 50%-68% mean absolute error reduction than the current state-of-the-art system trained for news sentence specificity. We also demonstrate the potential of our work in improving the quality and informativeness of dialogue generation systems.

Reducing Uncertainty in EQ-5D Value Sets: The Role of Spatial Correlation.

Scoring algorithms for the EQ-5D-3L are constructed subject to a large degree of uncertainty (a credible interval width of 0.152, which is significant in comparison to the reported minimal important differences). The purpose of this work is to explore modeling techniques that will reduce the extent of this uncertainty. We used the US valuation study data. A Bayesian approach was used to calculate predicted utilities and credible intervals. A spatial Gaussian correlation structure was used to model correlation among health states (HS), thus allowing directly valued HS to contribute to the predicted utility of nearby unvalued HS. Leave-one-out cross-validation was used to compare model performances. The average posterior standard deviation was 0.039 for the unvalued health states and 0.011 for the valued health states. Using cross-validation, the US D1 model had 31% coverage probability. Models with independent and Gaussian correlation had coverage probabilities of 95% and 93%, respectively. Moreover, the Gaussian correlation structure resulted in a 25.6% reduction in mean squared error (SE) and 13.2% reduction in mean absolute error (AE) compared to the independent correlation structure (mean SE of 0.00131 v. 0.00176 and mean AE of 0.02818 v. 0.03248). Uncertainty was substantially lower for the directly valued HS compared to unvalued HS, suggesting direct valuation of as many health states as possible. Incorporation of a spatial correlation significantly reduced uncertainty. Hence, we suggest incorporating this when constructing scoring algorithms.

Improving Visual Field Examination of the Macula Using Structural Information.

PurposeTo investigate a novel approach for structure-function modeling in glaucoma to improve visual field testing in the macula.MethodsWe acquired data from the macular region in 20 healthy eyes and 31 with central glaucomatous damage. Optical coherence tomography (OCT) scans were used to estimate the local macular ganglion cell density. Perimetry was performed with a fundus-tracking device using a 10-2 grid. OCT scans were matched to the retinal image from the fundus perimeter to accurately map the tested locations onto the structural damage. Binary responses from the subjects to all presented stimuli were used to calculate the structure-function model used to generate prior distributions for a ZEST (Zippy Estimation by Sequential Testing) Bayesian strategy. We used simulations based on structural and functional data acquired from an independent dataset of 20 glaucoma patients to compare the performance of this new strategy, structural macular ZEST (MacS-ZEST), with a standard ZEST.ResultsCompared to the standard ZEST, MacS-ZEST reduced the number of presentations by 13% in reliable simulated subjects and 14% with higher rates (≥20%) of false positive or false negative errors. Reduction in mean absolute error was not present for reliable subjects but was gradually more important with unreliable responses (≥10% at 30% error rate).ConclusionsBinary responses can be modeled to incorporate detailed structural information from macular OCT into visual field testing, improving overall speed and accuracy in poor responders.Translational RelevanceStructural information can improve speed and reliability for macular testing in glaucoma practice.

Neural network cloud top pressure and height for MODIS

Abstract. Cloud top height retrieval from imager instruments is important for nowcasting and for satellite climate data records. A neural network approach for cloud top height retrieval from the imager instrument MODIS (Moderate Resolution Imaging Spectroradiometer) is presented. The neural networks are trained using cloud top layer pressure data from the CALIOP (Cloud-Aerosol Lidar with Orthogonal Polarization) dataset. Results are compared with two operational reference algorithms for cloud top height: the MODIS Collection 6 Level 2 height product and the cloud top temperature and height algorithm in the 2014 version of the NWC SAF (EUMETSAT (European Organization for the Exploitation of Meteorological Satellites) Satellite Application Facility on Support to Nowcasting and Very Short Range Forecasting) PPS (Polar Platform System). All three techniques are evaluated using both CALIOP and CPR (Cloud Profiling Radar for CloudSat (CLOUD SATellite)) height. Instruments like AVHRR (Advanced Very High Resolution Radiometer) and VIIRS (Visible Infrared Imaging Radiometer Suite) contain fewer channels useful for cloud top height retrievals than MODIS, therefore several different neural networks are investigated to test how infrared channel selection influences retrieval performance. Also a network with only channels available for the AVHRR1 instrument is trained and evaluated. To examine the contribution of different variables, networks with fewer variables are trained. It is shown that variables containing imager information for neighboring pixels are very important. The error distributions of the involved cloud top height algorithms are found to be non-Gaussian. Different descriptive statistic measures are presented and it is exemplified that bias and SD (standard deviation) can be misleading for non-Gaussian distributions. The median and mode are found to better describe the tendency of the error distributions and IQR (interquartile range) and MAE (mean absolute error) are found to give the most useful information of the spread of the errors. For all descriptive statistics presented MAE, IQR, RMSE (root mean square error), SD, mode, median, bias and percentage of absolute errors above 0.25, 0.5, 1 and 2 km the neural network perform better than the reference algorithms both validated with CALIOP and CPR (CloudSat). The neural networks using the brightness temperatures at 11 and 12 µm show at least 32 % (or 623 m) lower MAE compared to the two operational reference algorithms when validating with CALIOP height. Validation with CPR (CloudSat) height gives at least 25 % (or 430 m) reduction of MAE.

Intercomparison of three methods to estimate evapotranspiration over temperate meadow in Inner Mongolia: Penman–Monteith, Makkink and Priestley–Taylor equation

AbstractThree methods of calculating evapotranspiration (ET), Penman–Monteith (PM), Makkink (MK) and Priestley–Taylor (PT) equation, were applied to a temperate meadow in Inner Mongolia, China. We compared these methods with eddy covariance (EC) method. The results indicated that a reduction in mean absolute error (MAE) and root mean squared error (RMSE) was achieved for MK and PT methods (45.6–83.0%) when using site‐specific parameters compared with referenced ones, but an increase of 75.9–93.9% for PM method. Intercomparison of three methods through comparison with EC measurements indicated that PT method with site‐specific parameters performed the best to estimate ET, followed by PM method with referenced parameters, and then MK equation with site‐specific parameters (overestimate ET by 8.2, 12.1, 16.5% and 3.4, 12.8, 21.0% respectively for half‐hourly and daily values).

PowKMeans

This article describes how recommender systems are software applications or web portals that generate personalized preferences using information filtering techniques, with a goal to support decision-making of the users. Collaborative-based techniques are often used to predict the unknown preferences of the user based upon his past preferences or the preferences of the similar users that have already been identified. A user which has a high correlation with any group of users is known as white user whereas the users which have less correlation with any group of users are known as gray-sheep users. The presence of gray-sheep users affects the accuracy of the model, and generates inaccurate predictions. To improve the prediction accuracy, it is important to differentiate graysheep users from white users. Experimental results show that PowKMeans is effective in improving the prediction accuracy by 4.62%. It has also shown reduction in Mean Absolute Error by 0.7757.

Sparse characterization of PPG based on K- SVD for beat-to-beat blood pressure prediction

Continuous non-invasive monitoring of blood pressure is essential for cardiovascular risk patients. This article presents the sparse characterization of Photoplethysmogram (PPG) using K-SVD technique for beat-to-beat blood pressure estimation. The relative changes in the blood volume that is represented by PPG is influenced by the pressure changes in the peripheral circulation. Owing to the anomalous nature of the time domain features obtained from ECG and PPG, we propose the use of sparse representation as means of feature extraction method. The sparse features generated from K-SVD processed dictionary that can approximate the shape of the PPG signal were taken as the features to predict the BP values. The proposed system is evaluated using the Multi-Parameter Intelligent Monitoring for Intensive Care (MIMIC-II) database. The proposed method is compared with the baseline system that employs time domain features for BP prediction. The Mean Absolute Error (MAE) and Root mean square Error (RMSE) between the predicted BP and ground truth BP were chosen as the performance measures. The system achieved the error measure of (MAE ± RMSE) 5.06 ± 6.27 mmHg for systolic BP and 2.99 ± 3.93 mm Hg for Diastolic BP. Further, the comparison studies suggests that the proposed system outperforms the baseline system with an overall reduction in MAE and RMSE by 20.55% and 22.14% respectively for systolic BP and 28.15% and 23.48% respectively for Diastolic BP. Hence, the sparse representation of PPG can be successfully utilized for the beat-to-beat prediction of BP.

MARSpline model for lead seven-day maximum and minimum air temperature prediction in Chennai, India

In this study, a Multivariate Adaptive Regression Spline (MARS) based lead seven days minimum and maximum surface air temperature prediction system is modelled for station Chennai, India. To emphasize the effectiveness of the proposed system, comparison is made with the models created using statistical learning technique Support Vector Machine Regression (SVMr). The analysis highlights that prediction accuracy of MARS models for minimum temperature forecast are promising for short term forecast (lead days 1 to 3) with mean absolute error (MAE) less than 1 °C and the prediction efficiency and skill degrades in medium term forecast (lead days 4 to 7) with slightly above 1 °C. The MAE of maximum temperature is little higher than minimum temperature forecast varying from 0.87 °C for day-one to 1.27 °C for lag day-seven with MARS approach. The statistical error analysis emphasizes that MARS models perform well with an average 0.2 °C of reduction in MAE over SVMr models for all ahead seven days and provide significant guidance for the prediction of temperature event. The study also suggests that the correlation between the atmospheric parameters used as predictors and the temperature event decreases as the lag increases with both approaches.

Estimating nonrigid motion from inconsistent intensity with robust shape features

To develop a nonrigid motion estimation method that is robust to heterogeneous intensity inconsistencies amongst the image pairs or image sequence. Intensity and contrast variations, as in dynamic contrast enhanced magnetic resonance imaging, present a considerable challenge to registration methods based on general discrepancy metrics. In this study, the authors propose and validate a novel method that is robust to such variations by utilizing shape features. The geometry of interest (GOI) is represented with a flexible zero level set, segmented via well-behaved regularized optimization. The optimization energy drives the zero level set to high image gradient regions, and regularizes it with area and curvature priors. The resulting shape exhibits high consistency even in the presence of intensity or contrast variations. Subsequently, a multiscale nonrigid registration is performed to seek a regular deformation field that minimizes shape discrepancy in the vicinity of GOIs. To establish the working principle, realistic 2D and 3D images were subject to simulated nonrigid motion and synthetic intensity variations, so as to enable quantitative evaluation of registration performance. The proposed method was benchmarked against three alternative registration approaches, specifically, optical flow, B-spline based mutual information, and multimodality demons. When intensity consistency was satisfied, all methods had comparable registration accuracy for the GOIs. When intensities among registration pairs were inconsistent, however, the proposed method yielded pronounced improvement in registration accuracy, with an approximate fivefold reduction in mean absolute error (MAE = 2.25 mm, SD = 0.98 mm), compared to optical flow (MAE = 9.23 mm, SD = 5.36 mm), B-spline based mutual information (MAE = 9.57 mm, SD = 8.74 mm) and mutimodality demons (MAE = 10.07 mm, SD = 4.03 mm). Applying the proposed method on a real MR image sequence also provided qualitatively appealing results, demonstrating good feasibility and applicability of the proposed method. The authors have developed a novel method to estimate the nonrigid motion of GOIs in the presence of spatial intensity and contrast variations, taking advantage of robust shape features. Quantitative analysis and qualitative evaluation demonstrated good promise of the proposed method. Further clinical assessment and validation is being performed.

Comparison of individual and combined ANN models for prediction of air and dew point temperature

Predicted air and dew point temperatures can be valuable in decision making in many areas including protecting crops from damage, avoiding heat stress on animals and humans, and in planning related to energy management. Current web-based artificial neural network (ANN) models on the Automated Environment Monitoring Network (AEMN) in Georgia predict hourly air and dew point temperature for twelve prediction horizons, using 24 models. The observed air temperature may approach the observed dew point temperature, but never goes below it. Current web based ANN models have prediction errors which, when the air and dew point temperatures are close, may cause air temperature to be predicted below the dew point temperature. Herein this error is referred to as a prediction anomaly. The goal of this research was to improve the prediction accuracy of existing air and dew point temperature ANN models by combining the two weather variables into a single ANN model for each prediction horizon. The objectives of this study were to reduce the mean absolute error (MAE) of prediction and to reduce the number of prediction anomalies. The combined models produced a reduction in the air temperature MAE for ten of twelve prediction horizons with an average reduction in MAE of 1.93 %. The combined models produced a reduction in the dew point temperature MAE for only six of twelve prediction horizons with essentially no average decrease in MAE. However, the combined models showed a marked reduction in prediction anomalies for all twelve prediction horizons with an average reduction of 34.1 %. The reduction in prediction anomalies ranged from 4.6 % at the one-hour horizon to 60.5 % at the eleven-hour horizon.

Nowcasting with INCA During SNOW-V10

A Central-European nowcasting system which has been developed for use in mountainous terrain is tested in the Whistler/Vancouver area as part of the SNOW-V10 experiment. The integrated nowcasting through comprehensive analysis system provides hourly updated gridded forecasts of temperature, humidity, and wind, as well as precipitation forecasts which are updated every 15 min. It is based on numerical weather prediction (NWP) output and real-time surface weather station and radar data. Verification of temperature, relative humidity, and wind against surface stations shows that forecast errors are significantly reduced in the nowcasting range compared to those of the driving NWP model. The main contribution to the improvement comes from the implicit bias correction due to use of the latest observations. Relative humidity shows the longest lasting effect, with >50 % reduction of mean absolute error up to +4 h. For temperature and wind speed this percentage is reached after +2 and +3 h, respectively. Two cases of precipitation nowcasting are discussed and verified qualitatively.

Maximum and Largest Crown Width Equations for 15 Tree Species in Maine

Abstract An extensive statewide data set for seven conifer and eight hardwood species commonly occurring in Maine was used in the development of maximum and largest crown width equations. To establish the characteristics of open-grown trees, quantile regression was used to estimate the biological maximum crown width for a species at a given diameter. To predict crown widths of trees in forested settings, a constrained nonlinear equation was used, using the predicted maximum crown width, tree diameter, and crown ratio. The models performed well across the wide range of stand conditions present in the data set and improved predictions over the currently used crown width equations for most species (reduction of mean absolute error ranged from 1 to 23%). In general, predictions of largest crown width were not greatly improved with the inclusion of crown ratio, and there was a high amount of unexplained variation for shade-tolerant hardwood species, such as American beech (Fagus grandifolia) and sugar maple (Acer saccharum). The equations presented herein can be used in examining tree crown profiles, computing measurements of stand density, and investigating canopy dynamics for species common to the forests of Maine.

Scale Filtering for Improved Nowcasting Performance in a High-Resolution X-Band Radar Network

Precipitation patterns exist on a continuum of scales where generally larger scale features facilitate longer useful prediction times at the expense of coarser resolution. Favorable measurement range and resolution make weather radar observations an attractive choice for input to automated short-term weather prediction (nowcasting) systems. Previous research has shown that nowcasting performance can be improved by spatially filtering radar observations and considering only those precipitation scales that are most representative of pattern motion for prediction or filtering those scales from predicted fields deemed unpredictable by remaining past their lifetimes. It has been shown that an improvement in nowcasting performance can be obtained by first applying a nonlinear elliptical spatial filter to observed Weather Surveillance Radar 88 Doppler vertically integrated liquid water fields to predict motion of larger scale features believed to better represent the motion of the entire precipitation pattern for forecast lead times up to 1 h. It has also been shown in the literature that wavelet transform can be used to develop measures of predictability at each scale and adaptive wavelet filters can be designed to remove perishable scales from predicted continental-scale reflectivity fields according to prediction lead time. This paper investigates the adaptation of both of these approaches and Fourier filtering to evaluate the effects of scale filtering on nowcasting performance using a Fourier-based nowcasting method and high-resolution Engineering Research Center for Collaborative Adaptive Sensing of the Atmosphere radar reflectivity data. A maximum improvement of approximately 18% in terms of Critical Success Index was observed by applying Fourier filtering in the context of truncating Fourier coefficients within the prediction model to the observed sequence of reflectivity fields used for assimilation. In addition, applying Fourier filtering to the resulting predictions showed a maximum reduction in mean absolute error of approximately 14%.

Isotope thermometry in melt‐affected ice cores

A statistically significant relationship is observed between stable water isotopes (δ18O) and melt amounts in a melt‐affected firn core (SSummit) taken from the Prince of Wales Icefield, Ellesmere Island, Canada. By contrast, a low‐melt firn core taken from a higher‐elevation, higher‐latitude location on the same icefield shows no relationship between these variables. We interpret this as evidence for meltwater‐induced isotopic enrichment at SSummit. A percent melt–based correction slope is applied to isotopic values from SSummit. Uncorrected and corrected temperature records derived from the raw and corrected δ18O values are compared to bias‐corrected temperature data from the NCEP Reanalysis. Improvements are observed in the isotopic reconstruction of SSummit annual precipitation‐weighted temperatures when we correct for meltwater enrichment, with a reduction from +0.6°C to 0.0°C in the mean annual error and a decrease in root‐mean‐square error from 1.8°C to 1.6°C. The correction factor appears to overcorrect isotopic modification during high melt years such as 1999, during which SSummit experienced nearly 70% more melt than the average from 1975 to 2000. Excluding 1999 data from the correction analysis results in a slight reduction in mean absolute error from 1.4°C to 1.3°C. These results suggest that melt‐induced isotopic modification cannot be corrected in very high melt years.

Modeling and forecasting the mean hourly wind speed time series using GMDH-based abductive networks

Wind speed forecasts are important for the operation and maintenance of wind farms and their profitable integration into power grids, as well as many important applications in shipping, aviation, and the environment. Modern machine learning techniques including neural networks have been used for this purpose, but it has proved hard to make significant improvements on the performance of the simple persistence model. As an alternative approach, we propose here the use of abductive networks, which offer the advantages of simplified and more automated model synthesis and transparent analytical input–output models. Various abductive models for predicting the mean hourly wind speed 1 h ahead have been developed using wind speed data at Dhahran, Saudi Arabia during the month of May over the years 1994–2005. The models were evaluated on the data for May 2006. Models described include a single generic model to forecast next-hour speed from the previous 24 hourly measurements and an hour index, which give an overall mean absolute error (MAE) of 0.85 m/s and a correlation coefficient of 0.83 between actual and predicted values. The model achieves an improvement of 8.2% reduction in MAE compared to hourly persistence. The above model was used iteratively to forecast the hourly wind speed 6 h and 24 h ahead at the end of a given day, with MAEs of 1.20 m/s and 1.42 m/s which are lower than forecasting errors based on day-to-day persistence by 14.6% and 13.7%. Relative improvements on persistence exceed those reported for several machine learning approaches reported in the literature.

Reduction of halftoning errors due to line load in a plasma display panel

The displayed gray level of a plasma display panel (PDP) is often different from the intended target gray level because the displayed level is affected by line load (the number of on-cells in a scan line). Previous attempts to reduce this difference (i.e. error) have been applicable to PDPs with no constraint imposed on the on-off decision for each sub-field. In order to improve the image quality (e.g. dynamic false contour (DFC) reduction), recently-developed PDPs impose constraints on sub-field coding and reduce the number of gray levels. This paper proposes a new sub-field coding algorithm that compensates for level error caused by the load when constraints are imposed on sub-field coding. As the number of gray levels is reduced, halftoning is employed in order to represent intermediate gray levels. In the proposed algorithm, the magnitude of the level error caused by the load effect is predicted and then two levels that are used for the halftoning are modified to compensate for the error. The modified levels are the nearest levels to the input level among the levels that satisfy the constraint of the sub-field coding. In order to reduce errors with moderate computational complexity, error prediction and level update are repeated by twice. A significant error reduction is achieved and shown by simulation and experimentation using a 42-inch PDP, including a 93.5% reduction in the mean absolute error (MAE) in simulations and a 67.2% reduction in MAE in experiments on a PDP.

Branch characteristics of widely spaced Douglas-fir in south-western Germany: Comparisons of modelling approaches and geographic regions

Models of Douglas-fir branch and whorl characteristics were developed from contrasting spacing experiments in southwest Germany. The dataset was based on 100 young (20–30 years old), unpruned and partially pruned trees from a 100, 200, and 1200 stems ha −1 spacing experiment on Douglas-fir that was replicated 3 times across the region. The material was used to predict (1) the number of branches whorl −1, (2) branch angle, (3) status (living/dead) of the branches within the living crown, (4) maximum branch diameter whorl −1, and (5) relative diameter of branches within a whorl. For each of these models (except branch status), both a linear and nonlinear, generalised hierarchical mixed effects equation was developed. The comparison of the linear and nonlinear approaches showed that both had a relatively similar level of bias, but the nonlinear equations generally performed better (reduction in mean absolute error of 1.1–69.5%). Overall, individual branch and tree properties were sufficient to give logical and precise predictions of the branch characteristics for the models across the range of sampled stand densities. In addition, the models showed a similar behaviour compared to models on Douglas-fir crown structure from the Pacific Northwest, USA. This suggests that the allometric relationship between tree size and branch characteristics for a given species may be relatively consistent across regions, even ones with highly contrasting growing conditions like in this study. The models performed well across a range of stand conditions and now will be further integrated into an individual tree growth and yield simulations system.

A statistical method for forecasting extreme daily temperatures using ECMWF 2‐m temperatures and ground station measurements

AbstractA recursive statistical method is given to post‐process the 2‐metre temperature ECMWF General Circulation Model by using the SAR network of surface stations in Sardinia. The method calculates daily maximum and minimum temperature forecasts up to 5 days ahead for 55 stations. Forecasts over one year for 55 ground stations have been analysed. A comparison with Direct Model Output (DMO) and persistence is then made, giving special attention to the winter season and to freezing conditions. The error analysis shows a BIAS very close to zero (between 0°C and 0.5°C) on all the stations and a marked reduction of mean absolute error compared to the DMO forecast. The error reduction is more marked in the case of freezing conditions than with the DMO forecast. Such situations are difficult to forecast and the very high BIAS of DMO has not been eliminated completely. The advantages of the procedure are: (a) there is no need to store large numbers of measurements and large amounts of model forecast data; (b) the estimated correction is easily adapted to new meteorological conditions; (c) the procedure is transparent to the updating of the model; and (d) the procedure is very easy to implement operationally. Copyright © 2004 Royal Meteorological Society

Beat-to-beat noise removal in noninvasive His-bundle electrocardiogram

His-bundle electrocardiogram micropotentials are usually obtained by serial-signal averaging, because no acceptable, satisfactory solution for beat-by-beat removal of power-line interference and electromyographic noise has been found. A method has been developed for surface beat-to-beat His-bundle potential recovery, with the hypothesis that no distortion in the signal shape is admissible. It is based on consecutive power-line interference subtraction and wavelet-domain electromyographic noise suppression, modified to match the strict criteria for detecting low-amplitude His potentials. The beat-to-beat wavelet-domain Wiener filtering uses a pilot signal estimate obtained from a limited number (around 20) of heart beat averages. The method resulted in an improvement of more than 4.5 dB in the signal-to-noise ratio and more than 20% reduction in mean absolute error, both measured along the P-Q segment. It is applicable for ECG signals contaminated by moderate electromyogram noise, with an initial signal-to-noise ratio of 15 dB or higher.