1. Introduction

2. Materials and Methods

2.1. Dataset Description

In this study, the Massachusetts Institute of Technology-Beth Israel Hospital (MIT-BIH) Arrhythmia Dataset [15] is utilized. This dataset includes 24-hour ECG recordings collected from 47 individuals, with subjects ranging in age from 23 to 89 years. We utilized excerpts from 30-minute, dual-channel ECG recordings collected from 47 participants between 1975 and 1979. The dataset features two ECG lead signals for each recording, digitized at 360 samples per second with 11-bit resolution over a 10-mV range.

For the purposes of classification, the annotations were grouped into five categories following the AAMI EC57 standard [16]. Class N encompasses "normal beats," as well as "left bundle branch block," "right bundle branch block," "atrial escape," and "nodal junction escape" beats. Class V includes "premature ventricular contractions" and "ventricular escape" beats. Class S contains "atrial premature" (AP) beats, "aberrated premature" beats, "nodal junction premature" beats, and "supraventricular" beats. Class Q, sometimes referred to as "unknown beats," includes "paced beats" (P), "fusion of paced and normal beats" (fPN), and "unclassified beats." Lastly, Class F is comprised of "fusion of ventricular and normal beats" (fVN). These annotations are derived from specific symbols present in the original dataset, which were mapped into the five simplified categories for classification purposes. For example, the "N" label includes symbols such as ’N’, ’L’, ’R’, ’e’, and ’j’, while the "S" label corresponds to symbols like ’A’, ’a’, and ’S’. This mapping was implemented to standardize the labels for the classification task.

For the purposes of classification, the annotations were grouped into five categories following the AAMI EC57 standard [16]. Class N encompasses "normal beats," as well as "left bundle branch block," "right bundle branch block," "atrial escape," and "nodal junction escape" beats. Class V includes "premature ventricular contractions" and "ventricular escape" beats. Class S contains "atrial premature" (AP) beats, "aberrated premature" beats, "nodal junction premature" beats, and "supraventricular" beats. Class Q, sometimes referred to as "unknown beats," includes "paced beats" (P), "fusion of paced and normal beats" (fPN), and "unclassified beats." Lastly, Class F is comprised of "fusion of ventricular and normal beats" (fVN). These annotations are derived from specific symbols present in the original dataset, which were mapped into the five simplified categories for classification purposes. For example, the "N" label includes symbols such as ’N’, ’L’, ’R’, ’e’, and ’j’, while the "S" label corresponds to symbols like ’A’, ’a’, and ’S’. This mapping was implemented to standardize the labels for the classification task. The illustration of each signals were shown in Figure 2.

The classification and understanding of these beats rely heavily on the identification of the PQRST complex, which is fundamental to interpreting ECG signals. The PQRST complex consists of five main components: the P wave, which indicates atrial depolarization; the QRS complex, which represents ventricular depolarization; and the T wave, which denotes ventricular repolarization. Occasionally, a U wave may be present, following the T wave, whose origin is less clearly understood but is thought to be related to the repolarization of the Purkinje fibers. Figure 1 provides a visual representation of these components within a typical ECG waveform. Understanding the morphology and timing of these components is critical for accurate beat classification and diagnosis, as each class of beats exhibits distinct characteristics in terms of the PQRST sequence. For instance, ventricular contractions such as those in Class V may show an abnormal QRS complex, while atrial contractions in Class S alter the P wave’s morphology.

Figure 1. Illustration of each label in an ECG signal.

Figure 1. Illustration of each label in an ECG signal.

Figure 2. Illustration of time series signals for each label.

Figure 2. Illustration of time series signals for each label.

2.4. 1D-CNN Model Architecture

A Convolutional Neural Network (CNN) is a type of deep learning architecture that excels at extracting high-level features from input data. Conceptually similar to a multilayer perceptron (MLP), each neuron in a CNN has an activation function that processes weighted inputs to produce outputs. CNNs are composed primarily of three types of layers: convolutional layers, pooling layers, and fully connected layers [21]. These networks, when adequately trained, find application in various fields such as speech recognition, structural engineering, and image processing.

A 1D-CNN is a variant of the traditional CNN, tailored specifically for handling one-dimensional data, making it particularly effective for sparse datasets that traditional CNNs might struggle with. Unlike 2D-CNNs that use two-dimensional convolutional filters, 1D-CNN utilize one-dimensional filters to extract features from data. This makes 1D-CNN not only more suitable for processing data such as audio or text but also more computationally efficient due to having fewer parameters. Their design allows them to capture local features within the signal, making them robust to slight shifts in time, which is crucial for analyzing time-series data with temporal components.

The first 1D-CNN classification model used in this study is designed to process the ECG signals and classify them into one of the five heartbeat categories. The architecture, as shown in Table, consists of:

• Three convolutional layers with filter sizes of 32, 64, and 64 respectively, and a kernel size of 3. Each convolutional layer is followed by a max pooling layer with a pool size of 2 to reduce the spatial dimensions of the data.
• Flattening layer, which transforms the 1D convoluted data into a flat vector for the fully connected layers.
• Two dense layers: the first dense layer has 64 units, followed by another dense layer with 32 units, both using the ReLU activation function.
• Dropout layer (with a dropout rate of 0.5) is added after the first dense layer to prevent overfitting by randomly deactivating neurons during training.
• The final output layer uses a softmax activation function to predict the probability of each class, with 5 output units corresponding to the five heartbeat categories.

This architecture was chosen for its ability to capture local patterns in the ECG signals while maintaining a manageable number of parameters. The model was trained using the Adam optimizer, with categorical cross-entropy as the loss function since this is a multi-class classification problem. The model was trained for 15 epochs with a batch size of 32, in line with prior research settings. The training dataset was split into 60% training and 40% testing to ensure the model was evaluated on a substantial portion of the data. Validation data was taken from the test set to monitor the model’s performance during training. Additionally, the class distribution in the training and test sets was adjusted to reflect the balanced dataset achieved through SMOTE.

Table 1. Parameter Tuning of the Proposed 1D CNN Model.

Table 1. Parameter Tuning of the Proposed 1D CNN Model.

Parameter	Value
Pooling Type	Max Pooling
Pooling Size	2
Units in First Dense Layer	64
Units in Second Dense Layer	32
Activation Function	ReLU
Dropout Rate	0.5
Output Layer Activation	Softmax
Number of Output Units	5
Optimizer	Adam

3. Results

3.2. Performance of Deep Learning Models on Preprocessed Signals

Confusion Matrix Analysis In classifying ECG signals using different windowing techniques in the preprocessing, the evaluation of classification performance is central, with a specific focus on confusion matrices for the Blackman, Hamming, and Hann windows. Classification matrix is a crucial metric to assess how well the classification algorithms perform under the different spectral conditions induced by the respective window types.

The confusion matrix for the Blackman window exhibits prominent diagonal elements, with a notable 10,467 correct classifications for class F, indicating strong accuracy in classification across the observed categories. Misclassifications are minimal, with the most significant confusion between classes F and V, consisting of 35 instances, which is relatively low in comparison to the total classifications. This suggests that the Blackman window maintains high specificity and sensitivity, particularly in distinguishing between closely related classes.

Similarly, the Hamming window’s confusion matrix shows high values on the diagonal, reflective of effective classification, with class F having 10,458 correct classifications. However, this window displays a broader spread of misclassifications, particularly between classes F and V (38 instances), and between classes S and V (40 instances). These figures suggest a slightly reduced specificity under the Hamming window, which might influence its suitability in applications requiring high precision in class differentiation.

The Hann window’s matrix also demonstrates strong diagonal values, albeit with a slight reduction in sensitivity for class S, which has 2,361 correct classifications—a decrease compared to the other windows. Misclassifications are higher, especially between classes F and S (181 instances), and between classes S and V (44 instances), indicating potential challenges in distinguishing these class pairs more so than with the other window functions.

Overall, the comparison of these three window functions reveals robust classification capabilities across all types, yet highlights subtle differences in their performance. The Blackman window appears to offer the best overall accuracy with the least amount of misclassification, making it potentially more suitable for clinical applications where precise class distinction is crucial. On the other hand, the Hamming and Hann windows, while still effective, exhibit a slightly increased tendency for misclassification between certain classes, which could influence their application in specific diagnostic settings.

Overall Performance The performance analysis of the three window functions— Hamming, Hann, and Blackman—on ECG signal classification provides a detailed insight into their effectiveness across five classes (F, N, S, V, Q), evaluated through metrics such as Precision, Recall, and F1-Score ( Table 2). These metrics are defined as follows:

• Precision: The ratio of correctly predicted positive observations to the total predicted positive observations.

  $$\mathrm{Precision}={\displaystyle \frac{TP}{TP+FP}}$$

  where $TP$ is the number of true positives, and $FP$ is the number of false positives.
• Recall: The ratio of correctly predicted positive observations to all observations in the actual class.

  $$\mathrm{Recall}={\displaystyle \frac{TP}{TP+FN}}$$

  where $FN$ is the number of false negatives.
• F1-Score: The harmonic mean of Precision and Recall, providing a balance between the two.

  $$\mathrm{F}1-\mathrm{Score}=2\times {\displaystyle \frac{\mathrm{Precision}\times \mathrm{Recall}}{\mathrm{Precision}+\mathrm{Recall}}}$$

All three models demonstrate excellent performance in classifying class F, with notably high precision and recall, suggesting their strong capability in accurately identifying this category with minimal false positives and negatives. Classes N and S also see relatively good performance, although the recall for class N is slightly reduced, indicating fewer true positives detected relative to the actual positives present. This could impact the clinical utility of the model where missing a positive case can have significant consequences.

Figure 4. Comparison of confusion matrix for ECG signals classification processed using (a) Hann, (b) Hamming, and (c) Blackman windows.

Figure 4. Comparison of confusion matrix for ECG signals classification processed using (a) Hann, (b) Hamming, and (c) Blackman windows.

The performance on class V is moderate across all models, with precision slightly lower than other classes, hinting at a higher rate of false positives for this class. This might suggest a common challenge in the classification of this class using window-based spectral analysis, possibly due to the overlap in spectral characteristics between class V and other classes.

A notable observation across all three models is the complete lack of detection of class Q, which exhibits zero values in precision, recall, and F1-Score. This uniform failure to detect class Q likely indicates that class Q represents a type of signal that is undetectable under the current feature extraction and classification methodology employed. This could be due to the absence of distinct or adequate features within the spectral domain that these window functions analyze, or possibly the absence of sufficient representative samples of class Q in the training dataset. Overall, while all three window functions provide robust classification capabilities for most ECG signal classes, the choice among them should consider the specific classification needs and the unique challenges presented by each class. The slight differences in performance metrics between the models can guide the selection of a window function based on whether higher precision or a better-balanced F1-Score is more critical for the intended application.

The results clearly demonstrate the utility of FIR window functions in enhancing the signal-to-noise ratio and reducing artifacts in ECG signals, which in turn significantly improves the performance of deep learning models. The use of Blackman window filtering, in particular, facilitated a more refined feature extraction process, thereby aiding in more accurate heart condition diagnostics. The findings suggest that integrating advanced FIR filtering techniques with deep learning frameworks can significantly advance the field of biomedical signal analysis.

4. Conclusion

References

1. Antzelevitch, C.; Burashnikov, A. Overview of Basic Mechanisms of Cardiac Arrhythmia, 2011. [CrossRef]
2. Varalakshmi, P.; Sankaran, A.P. An improved hybrid AI model for prediction of arrhythmia using ECG signals. Biomedical Signal Processing and Control 2023, 80. [Google Scholar] [CrossRef]
3. Xiao, Q.; Lee, K.; Mokhtar, S.A.; Ismail, I.; bin Md Pauzi, A.L.; Zhang, Q.; Lim, P.Y. Deep Learning-Based ECG Arrhythmia Classification: A Systematic Review, 2023. [CrossRef]
4. Ahmed, A.A.; Ali, W.; Abdullah, T.A.; Malebary, S.J. Classifying Cardiac Arrhythmia from ECG Signal Using 1D CNN Deep Learning Model. Mathematics 2023, 11. [Google Scholar] [CrossRef]
5. Eleyan, A.; Alboghbaish, E. Multi-Classifier Deep Learning based System for ECG Classification Using Fourier Transform. Institute of Electrical and Electronics Engineers Inc., 2023. [CrossRef]
6. Ullah, H.; Heyat, M.B.B.; Akhtar, F.; Sumbul.; Muaad, A.Y.; Islam, M.S.; Abbas, Z.; Pan, T.; Gao, M.; Lin, Y.; Lai, D. An End-to-End Cardiac Arrhythmia Recognition Method with an Effective DenseNet Model on Imbalanced Datasets Using ECG Signal. Computational Intelligence and Neuroscience 2022, 2022. [CrossRef]
7. Zhang, H.; Liu, C.; Zhang, Z.; Xing, Y.; Liu, X.; Dong, R.; He, Y.; Xia, L.; Liu, F. Recurrence Plot-Based Approach for Cardiac Arrhythmia Classification Using Inception-ResNet-v2. Frontiers in Physiology 2021, 12. [Google Scholar] [CrossRef] [PubMed]
8. Yeh, Y.C.; Chiou, C.W.; Lin, H.J. Analyzing ECG for cardiac arrhythmia using cluster analysis. Expert Systems with Applications 2012, 39, 1000–1010. [Google Scholar] [CrossRef]
9. Dhyani, S.; Kumar, A.; Choudhury, S. Analysis of ECG-based arrhythmia detection system using machine learning. MethodsX 2023, 10. [Google Scholar] [CrossRef] [PubMed]
10. Śmigiel, S.; Pałczyński, K.; Ledziński, D. ECG signal classification using deep learning techniques based on the PTB-XL dataset. Entropy 2021, 23. [Google Scholar] [CrossRef] [PubMed]
11. Ansari, Y.; Mourad, O.; Qaraqe, K.; Serpedin, E. Deep learning for ECG Arrhythmia detection and classification: an overview of progress for period 2017–2023, 2023. [CrossRef]
12. Aziz, S.; Ahmed, S.; Alouini, M.S. ECG-based machine-learning algorithms for heartbeat classification. Scientific Reports 2021, 11. [Google Scholar] [CrossRef] [PubMed]
13. Biran, A.; Jeremic, A. ECG based Human Identification using Short Time Fourier Transform and Histograms of Fiducial QRS Features. SciTePress, 2020, pp. 324–329. [CrossRef]
14. Kumar, M.A.; Chakrapani, A. Classification of ECG signal using FFT based improved Alexnet classifier. PLoS ONE 2022, 17. [Google Scholar] [CrossRef] [PubMed]
15. Moody, G.B.; Mark, R.G. MIT-BIH Arrhythmia Database, 1992. [CrossRef]
16. Yang, M.; Liu, W.; Zhang, H. A robust multiple heartbeats classification with weight-based loss based on convolutional neural network and bidirectional long short-term memory. Frontiers in Physiology 2022, 13. [Google Scholar] [CrossRef] [PubMed]
17. Bracewell, R.N.; Bracewell, R.N. The Fourier transform and its applications; Vol. 31999, McGraw-Hill New York, 1986.
18. Oppenheim, A.V. Discrete-time signal processing; Pearson Education India, 1999.
19. Podder, P.; Khan, T.Z.; Khan, M.H.; Rahman, M.M. Comparative Performance Analysis of Hamming, Hanning and Blackman Window, 2014.
20. Kaur, M.; Kaur, S.P. High Frequency Noise Removal From Electrocardiogram Using Fir Low Pass Filter Bassed On Window Technique 2018. 8, 27–32.
21. Kiranyaz, S.; Avci, O.; Abdeljaber, O.; Ince, T.; Gabbouj, M.; Inman, D.J. 1D convolutional neural networks and applications: A survey. Mechanical Systems and Signal Processing 2021, 151, 107398. [Google Scholar] [CrossRef]
22. Czanner, G.; Sarma, S.V.; Ba, D.; Eden, U.T.; Wu, W.; Eskandar, E.; Lim, H.H.; Temereanca, S.; Suzuki, W.A.; Brown, E.N. Measuring the signal-to-noise ratio of a neuron. Proceedings of the National Academy of Sciences of the United States of America 2015, 112, 7141–7146. [Google Scholar] [CrossRef] [PubMed]