Abstract:Modulation classification, an intermediate process between signal detection and demodulation in a physical layer, is now attracting more interest to the cognitive radio field, wherein the performance is powered by artificial intelligence algorithms. However, most existing conventional approaches pose the obstacle of effectively learning weakly discriminative modulation patterns. This paper proposes a robust modulation classification method by taking advantage of deep learning to capture the meaningful information of modulation signal at multi-scale feature representations. To this end, a novel architecture of convolutional neural network, namely Chain-Net, is developed with various asymmetric kernels organized in two processing flows and associated via depth-wise concatenation and element-wise addition for optimizing feature utilization. The network is evaluated on a big dataset of 14 challenging modulation formats, including analog and high-order digital techniques. The simulation results demonstrate that Chain-Net robustly classifies the modulation of radio signals suffering from a synthetic channel deterioration and further performs better than other deep networks.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is that in wireless communication, modulation classification, as an intermediate process between signal detection and demodulation, is becoming increasingly important in the field of cognitive radio. However, most of the existing traditional methods have obstacles in effectively learning weakly - discriminative modulation patterns. Specifically: 1. **Likelihood - Based Methods**: Although it can achieve optimal accuracy in an ideal situation, it performs poorly when dealing with continuous - phase modulation and consumes a large amount of computational resources. 2. **Feature - Based Methods**: Although it is simple to implement, has low complexity and strong robustness under synthetic channel degradation, it requires expertise in feature engineering and has weak capabilities when dealing with big data. To overcome these challenges, this paper proposes a robust modulation classification method based on deep learning, which captures meaningful information of modulation signals through multi - scale feature representation. To this end, the authors developed a new convolutional neural network architecture called Chain - Net. This network is organized into two processing streams through various asymmetric kernels and optimizes feature utilization through deep cascading and element - level addition. ### Main Contributions 1. **Novel Network Architecture**: Chain - Net optimizes feature utilization through multi - scale feature maps, and particularly performs well in dealing with modulation classification problems of high - order modulation formats and under heavy channel impairments. 2. **Large - scale Dataset**: The paper contributes a large - scale dataset containing 14 challenging modulation formats (including analog and digital techniques) for evaluating network performance. 3. **Performance Verification**: Simulation results show that the modulation classification performance of Chain - Net under synthetic channel degradation is better than other deep networks. ### Problems Solved - **Learning of Weakly - Discriminative Modulation Patterns**: Through multi - scale feature representation and deep learning, Chain - Net can learn the characteristics of modulation signals more effectively. - **Classification of High - Order Modulation Formats**: Especially under heavy channel impairments, Chain - Net can classify high - order modulation formats more accurately. - **Computational Efficiency**: Chain - Net maintains a low computational complexity while improving classification accuracy by optimizing the network structure. In conclusion, this paper aims to solve the deficiencies of existing modulation classification methods in dealing with complex modulation signals through deep learning methods, especially the robustness problem under synthetic channel degradation.

Chain-Net: Learning Deep Model for Modulation Classification Under Synthetic Channel Impairment

An Improved Modulation Recognition Algorithm Based on Fine-Tuning and Feature Re-Extraction

Learning Constellation Map with Deep CNN for Accurate Modulation Recognition

Automatic Modulation Classification: A Deep Architecture Survey

Deep Neural Network for Robust Modulation Classification Under Uncertain Noise Conditions

ChannelGAN: Deep Learning-Based Channel Modeling and Generating

Time and phase features network model for automatic modulation classification

Automatic Modulation Classification with Deep Neural Networks

Automatic Modulation Classification in Time-Varying Channels Based on Deep Learning

Enhanced Modulation Recognition Through Deep Transfer Learning in Hybrid Graph Convolutional Networks

Modulation Classification Through Deep Learning Using Resolution Transformed Spectrograms

Radio Modulation Classification Optimization Using Combinatorial Deep Learning Technique

An Efficient and Lightweight Model for Automatic Modulation Classification: A Hybrid Feature Extraction Network Combined with Attention Mechanism

A Novel Automatic Modulation Classification Scheme Based on Multi-Scale Networks

Reparameterization Causal Convolutional Network for Automatic Modulation Classification

Data-Transform Multi-Channel Hybrid Deep Learning for Automatic Modulation Recognition

Automatic Modulation Classification Using Hybrid Data Augmentation and Lightweight Neural Network

Automatic Modulation Recognition Based on a DCN-BiLSTM Network

Modulation Classification of VHF Communication System Based on CNN and Cyclic Spectrum Graphs

Deep multilevel architecture for automatic modulation classification

A IQ Correlation and Long-term Features Neural Network Classifier for Automatic Modulation Classification