Search papers, labs, and topics across Lattice.
This paper introduces Dualformer, a novel Transformer-based architecture that leverages a dual-channel neural network (DualNN) to efficiently process complex-valued signals for blind communication tasks like automatic modulation recognition (AMR) and signal scheme recognition (SSR). By sharing parameters between the real and imaginary components of the signals, Dualformer reduces generalization error while maintaining expressive capacity, outperforming traditional models in extensive experiments. The results show significant performance enhancements across multiple signal analysis tasks, highlighting the architecture's potential for broader applications in unsupervised and weakly supervised scenarios.
Dualformer achieves consistent performance improvements in complex-valued signal analysis, outperforming traditional methods by effectively sharing parameters across signal channels.
Designing effective feature extractors is critical for blind signal analysis tasks such as automatic modulation recognition (AMR), signal scheme recognition (SSR), and \color{black} signal structure parsing (SSP). In this work, we propose dual-channel neural network (DualNN) that efficiently exploits complex-valued signals through parameter sharing across IQ channels. Unlike traditional real-valued or complex-valued models, DualNN is a groundbreaking framework which shares the network parameters for processing the real and imaginary parts of the complex-valued signals, and is theoretically shown to reduce generalization error while preserving expressive capacity. Specifically, we propose a novel Transformer-based architecture to implement DualNN, called Dualformer. The Dualformer segments input signals into patch-level tokens and captures multi-granularity features, enabling robust performance across diverse signal analysis tasks. Furthermore, we conduct extensive experiments comparing Dualformer with three Transformer-based baselines and four conventional DL-based approaches. Results demonstrate consistent performance improvements on AMR, SSR, and SSP tasks. Besides, the modular design of DualNN allows it to generalize well to blind signal processing tasks such as blind source separation and low-SNR spectrum sensing. This work paves the way for a broader application of DualNN architectures in unsupervised and weakly supervised complex-valued signal analysis scenarios.
