Udit Satija

Real-Time Signal Quality-Aware ECG Telemetry System for IoT-Based Health Care Monitoring

In this paper, we propose a novel signal quality-aware Internet of Things (IoT)-enabled electrocardiogram (ECG) telemetry system for continuous cardiac health monitoring applications. The proposed quality-aware ECG monitoring system consists of three modules: 1) ECG signal sensing module; 2) automated signal quality assessment (SQA) module; and 3) signal-quality aware (SQAw) ECG analysis and transmission module. The main objectives of this paper are: design and development of a light-weight ECG SQA method for automatically classifying the acquired ECG signal into acceptable or unacceptable class and real-time implementation of proposed IoT-enabled ECG monitoring framework using ECG sensors, Arduino, Android phone, Bluetooth, and cloud server. The proposed framework is tested and validated using the ECG signals taken from the MIT-BIH arrhythmia and Physionet challenge databases and the real-time recorded ECG signals under different physical activities. Experimental results show that the proposed SQA method achieves promising results in identifying the unacceptable quality of ECG signals and outperforms existing methods based on the morphological and RR interval features and machine learning approaches. This paper further shows that the transmission of acceptable quality of ECG signals can significantly improve the battery lifetime of IoT-enabled devices. The proposed quality-aware IoT paradigm has great potential for assessing clinical acceptability of ECG signals in improvement of accuracy and reliability of unsupervised diagnosis system.

Performance study of cyclostationary based digital modulation classification schemes

Automatic Modulation Classification (AMC) is a essential component in Cognitive Radio (CR) for recognizing the modulation scheme. Many modulated signals manifest the property of cyclostationarity as a feature so it can be exploited for classification. In this paper, we study the performance of digital modulation classification technique based on the cyclostationary features and different classifiers such as Neural Network, Support Vector Machine, k-Nearest Neighbor, Naive Bayes, Linear Discriminant Analysis and Neuro-Fuzzy classifier. In this study we considered modulations i.e. BPSK, QPSK, FSK and MSK for classification. All classification methods studied using performance matrix including classification accuracy and computational complexity (time). The robustness of these methods are studied with SNR ranging from 0 to 20dB. Based upon the result we found that combining cyclostationary features with Naive Bayes and Linear Discriminant Analysis classifiers leads to provide better classification accuracy with less computational complexity.

Automatic modulation classification using S-transform based features

Automatic Modulation Classification plays a significant role in Cognitive Radio to identify the modulation format of the primary user. In this paper, we present the Stockwell transform (S-transform) based features extraction for classification of different digital modulation schemes using different classifiers such as Neural Network (NN), Support Vector Machine (SVM), Linear Discriminant Analysis (LDA), Naive Bayes (NB), k-Nearest Neighbor (k-NN). The S - transform provides time-frequency or spatial-frequency localization of a signal. This property of S-transform gives good discriminant features for different modulation schemes. Two simple features i.e., energy and entropy are used for classification. Different modulation schemes i.e., BPSK, QPSK, FSK and MSK are used for classification. The results are compared with wavelet transform based features using probability of correct classification, performance matrix including classification accuracy and computational complexity (time) for SNR range varying from 0 to 20 dB. Based upon the results, we found that S-transform based features outperform wavelet transform based features with better classification accuracy and less computational complexity.

A simple method for detection and classification of ECG noises for wearable ECG monitoring devices

An assessment of electrocardiogram (ECG) signal quality has become an unavoidable first step in most holter and ambulatory ECG signal analysis applications. In this paper, we present a simple method for automatically detection and classification of ECG noises. The proposed method consists of four major steps: moving average filter, blocking, feature extraction, and multistage decision-tree algorithm. In the proposed method, the dynamic amplitude range and autocorrelation maximum peak features are extracted for each block. In the first decision stage, a amplitude-dependent decision rule is used for detecting the presence of low-frequency (LF) noise (including, baseline wander (BW) and abrupt change (ABC) artifacts) and the high-frequency (HF) noise (including, power line interference (PLI) and muscle artifacts). In the second decision stage, the proposed method further classifies the LF noise into a BW noise or a ABC noise using the local dynamic amplitude range feature. The HF noise is classified into a PLI noise or a muscle noise using the local autocorrelation maximum peak feature. The proposed detection and classification method is tested and validated using a wide variety of clean and noisy ECG signals. Results show that the method can achieve an average sensitivity (Se) of 97.88%, positive productivity (+P) of 91.18% and accuracy of 89.06%.

A unified sparse signal decomposition and reconstruction framework for elimination of muscle artifacts from ECG signal

Removal of muscle artifacts from the ECG signals is crucial for a reliable and accurate measurement of local features of ECG signals. In this paper, we present an automatic method for removal of muscle artifacts from ECG signals, based on four steps: decomposing ECG signal using sparse signal decomposition on mixed dictionaries; obtaining QRS complex signal; determining time-instants of R-peak; and removal of muscle artifacts from ECG signal. The noise reduction performance of the proposed method is tested and validated using ECG signals taken from a standard MIT-BIH Arrhythmia database. The reconstructed signals are assessed using both subjective quality assessment test and objective quality assessment metrics. Performance evaluation results show that the proposed method outperforms other existing ECG denoising methods inadequately removing the muscle artifacts without significantly distorting the morphologies of P-wave, QRS-complex and T-wave of the ECG signals.

Digital modulation classification under non-Gaussian noise using sparse signal decomposition and maximum likelihood

In recent years, automatic signal detection and modulation classification play a vital role in the field of cognitive radio applications. The majority of the existing signals detection and classification methods assume that the received signal is contaminated by additive white Gaussian noise. Under impulsive noise condition, the performance of the traditional modulation classification methods may be degraded. Therefore, in this paper, we investigate the application of sparse signal decomposition using an overcomplete dictionary for detection and classification of digital modulation signals. The overcomplete hybrid dictionary consists of impulse waveform and sine and cosine waveform for effectively capturing morphological components of the impulse noise and deterministic modulated signals. The proposed modulation classification method includes the following steps: sparse signal decomposition (SSD) on hybrid dictionaries, modulated signal extraction, matched filtering, and maximum likelihood (ML) classification. The performance of the direct ML and SSD-based ML classification methods are tested and validated using different modulation techniques under different Gaussian and impulse noise conditions. The proposed system achieves a classification accuracy of 89 percent at 0 dB SNR and hence outperforms the direct ML method.

Noise-aware dictionary-learning-based sparse representation framework for detection and removal of single and combined noises from ECG signal

Automatic electrocardiogram (ECG) signal enhancement has become a crucial pre-processing step in most ECG signal analysis applications. In this Letter, the authors propose an automated noise-aware dictionary learning-based generalised ECG signal enhancement framework which can automatically learn the dictionaries based on the ECG noise type for effective representation of ECG signal and noises, and can reduce the computational load of sparse representation-based ECG enhancement system. The proposed framework consists of noise detection and identification, noise-aware dictionary learning, sparse signal decomposition and reconstruction. The noise detection and identification is performed based on the moving average filter, first-order difference, and temporal features such as number of turning points, maximum absolute amplitude, zerocrossings, and autocorrelation features. The representation dictionary is learned based on the type of noise identified in the previous stage. The proposed framework is evaluated using noise-free and noisy ECG signals. Results demonstrate that the proposed method can significantly reduce computational load as compared with conventional dictionary learning-based ECG denoising approaches. Further, comparative results show that the method outperforms existing methods in automatically removing noises such as baseline wanders, power-line interference, muscle artefacts and their combinations without distorting the morphological content of local waves of ECG signal.

Low-complexity detection and classification of ECG noises for automated ECG analysis system

Automated detection and classification of electrocardiogram (ECG) noise sources can play a crucial role in reliable measurement of ECG parameters for accurate diagnosis of cardiovascular diseases (CVDs) under unsupervised telehealth monitoring and intensive care unit (ICU) applications. Although the methods had quite acceptable detection rates, most methods are too complicated for real-time implementation for wearable cardiac health monitoring devices. Therefore, in this paper, we present a low-complexity algorithm for automatically detecting and classifying the ECG noises including flat line (FL), time-varying noise (TVN), baseline wander (BW), abrupt change (ABC), muscle artifact (MA) and power line interference (PLI). The proposed method is based on the moving average (MAv) and derivative filters and the five temporal features including turning points, global and local amplitude estimates, zerocrossing and autocorrelation peak values. The proposed method is tested and validated using a wide variety of clean and noisy ECG signals taken from the MIT-BIH arrhythmia database and Physionet Challenge database. The method achieves an average sensitivity (Se) of 98.55%, positive productivity (+P) of 95.27% and overall accuracy (OA) of 94.19% for classifying the noises. Results show that the proposed method outperforms the other existing methods.

Automated ECG Noise Detection and Classification System for Unsupervised Healthcare Monitoring

Objective: Automatic detection and classification of noises can play a vital role in the development of robust unsupervised electrocardiogram (ECG) analysis systems. This paper proposes a novel unified framework for automatic detection, localization, and classification of single and combined ECG noises. Methods : The proposed framework consists of the modified ensemble empirical mode decomposition (CEEMD), the short-term temporal feature extraction, and the decision-rule-based noise detection and classification. In the proposed framework, ECG signals are first decomposed using the modified CEEMD algorithm for discriminating the ECG components from the noises and artifacts. Then, the short-term temporal features such as maximum absolute amplitude, number of zerocrossings, and local maximum peak amplitude of the autocorelation function are computed from the extracted high-frequency and low-frequency signals. Finally, a decision rule-based algorithm is presented for detecting the presence of noises and classifying the processed ECG signals into six signal groups: noise-free ECG, ECG+BW, ECG+MA, ECG+PLI, ECG+BW+PLI, and ECG+BW+MA. Results: The proposed framework is rigorously evaluated on five benchmark ECG databases and the real-time ECG signals. The proposed framework achieves an average sensitivity of 99.12%, specificity of 98.56%, and overall accuracy of 98.90% in detecting the presence of noises. Classification results show that the framework achieves an average sensitivity, positive predictivity, and classification accuracy of 98.93%, 98.39%, and 97.38%, respectively. Conclusion: The proposed framework not only achieves better noise detection and classification rates than the current state-of-the-art methods but also accurately localizes short bursts of noises with low endpoint delineation errors. Significance: Extensive studies on benchmark databases demonstrate that the proposed framework is more suitable for reducing false alarm rates and selecting appropriate noise-specific denoising techniques in automated ECG analysis applications.

A robust sparse signal decomposition framework for baseline wander removal from ECG signal

In this paper, we present a new method based on sparse signal decomposition for effectively removing baseline wander (BW) noise from the ECG signal. Our proposed method not only eliminates BW noise from ECG signal but also preserves the morphological shape of local waves of the signal. Our proposed method is tested and validated on real ECG signals taken from MIT-BIH arrhythmia database. Comparative results depict the superior performance of the proposed method over the state of art methods.