Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
119 tokens/sec
GPT-4o
56 tokens/sec
Gemini 2.5 Pro Pro
43 tokens/sec
o3 Pro
6 tokens/sec
GPT-4.1 Pro
47 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

1D Convolutional Neural Networks and Applications: A Survey (1905.03554v1)

Published 9 May 2019 in eess.SP and cs.AI

Abstract: During the last decade, Convolutional Neural Networks (CNNs) have become the de facto standard for various Computer Vision and Machine Learning operations. CNNs are feed-forward Artificial Neural Networks (ANNs) with alternating convolutional and subsampling layers. Deep 2D CNNs with many hidden layers and millions of parameters have the ability to learn complex objects and patterns providing that they can be trained on a massive size visual database with ground-truth labels. With a proper training, this unique ability makes them the primary tool for various engineering applications for 2D signals such as images and video frames. Yet, this may not be a viable option in numerous applications over 1D signals especially when the training data is scarce or application-specific. To address this issue, 1D CNNs have recently been proposed and immediately achieved the state-of-the-art performance levels in several applications such as personalized biomedical data classification and early diagnosis, structural health monitoring, anomaly detection and identification in power electronics and motor-fault detection. Another major advantage is that a real-time and low-cost hardware implementation is feasible due to the simple and compact configuration of 1D CNNs that perform only 1D convolutions (scalar multiplications and additions). This paper presents a comprehensive review of the general architecture and principals of 1D CNNs along with their major engineering applications, especially focused on the recent progress in this field. Their state-of-the-art performance is highlighted concluding with their unique properties. The benchmark datasets and the principal 1D CNN software used in those applications are also publically shared in a dedicated website.

User Edit Pencil Streamline Icon: https://streamlinehq.com
Authors (6)
  1. Serkan Kiranyaz (86 papers)
  2. Onur Avci (12 papers)
  3. Osama Abdeljaber (3 papers)
  4. Turker Ince (22 papers)
  5. Moncef Gabbouj (167 papers)
  6. Daniel J. Inman (4 papers)
Citations (1,607)

Summary

A Survey on 1D Convolutional Neural Networks and Applications

This paper, authored by Serkan Kiranyaz et al., presents a comprehensive overview of 1D Convolutional Neural Networks (CNNs) and their significant engineering applications. 1D CNNs have been increasingly recognized as an effective alternative for various tasks traditionally dominated by 2D CNNs, especially in scenarios where the data primarily exists in a one-dimensional format, such as time series data. The authors explore the architecture, functioning, and practical implications of 1D CNNs, emphasizing their computational efficiency, adaptability, and real-time application potential.

Architectural Overview of 1D CNNs

1D CNNs share many similarities with their 2D counterparts, primarily in their use of convolutional layers for automatic feature extraction followed by fully connected layers for classification. However, they are specifically optimized for 1D data, utilizing linear convolutions which significantly reduce computational complexity. The authors detail the backpropagation process and the training methodology of 1D CNNs, highlighting how these networks can be efficiently trained with less computational power compared to 2D CNNs.

Key Engineering Applications

Real-time ECG Monitoring

The paper outlines the pioneering application of 1D CNNs in electrocardiogram (ECG) beat classification. This innovation allowed for real-time patient-specific cardiac arrhythmia detection, achieving accuracy rates of 99% for Ventricular Ectopic Beats and 97.6% for Supraventricular Ectopic Beats. A novel approach for early detection in healthy individuals, leveraging synthetic abnormal beats for training, marks a significant advancement in personalized cardiac health monitoring.

Structural Health Monitoring

In civil engineering, 1D CNNs have proven essential for structural damage detection and health monitoring. For example, in the Qatar University grandstand simulator, 1D CNNs were effective in detecting damages with high accuracy and real-time processing capabilities. The implementation with a wireless sensor network further demonstrated their practical applicability in large-scale infrastructure monitoring.

Rotating Machinery and Bearing Fault Detection

The paper discusses the use of 1D CNNs in the condition monitoring of rotating machinery, including bearing fault detection. In these applications, 1D CNNs have achieved near-perfect accuracy in identifying anomalies directly from the vibration signals, outperforming traditional methods both in accuracy and computational efficiency.

Fault Detection in Modular Multilevel Converters

In power electronics, 1D CNNs have been applied for real-time fault detection in Modular Multilevel Converters (MMCs). This application underscores the versatility of 1D CNNs in handling electrical signal anomalies, providing reliable and immediate fault diagnosis with minimal computational overhead.

Computational Complexity Analysis

The authors provide a detailed computational complexity analysis, illustrating that 1D CNNs require significantly fewer operations compared to 2D CNNs due to the simpler nature of their convolutional operations. This efficiency makes 1D CNNs particularly suited for applications requiring real-time processing on devices with limited computational resources.

Implications and Future Directions

The paper asserts that 1D CNNs offer a viable and efficient alternative for applications dealing with 1D signal data, particularly when labeled training data is limited or when real-time processing is essential. The authors suggest potential future developments in enhancing the neural architectures, possibly integrating more complex and heterogeneous neuron models as suggested by research into Generalized Operational Perceptrons (GOPs) and Operational Neural Networks (ONNs).

Conclusion

In summary, this survey underscores the robustness and applicability of 1D CNNs across various engineering fields. By combining feature extraction and classification into a singular learning process, 1D CNNs present a powerful tool for real-time, low-cost, and accurate signal processing. The continued evolution of these networks promises to extend their applicability and performance, paving the way for innovative solutions in both established and emerging domains.