Joint Signal Detection and Automatic Modulation Classification via Deep Learning

TL;DR

This work addresses the problem of jointly detecting multiple coexisting signals and classifying their modulations in a multi-carrier RF environment. It introduces CRML23, a synthetic dataset with random, multi-signal entries across a band, together with a joint detection and modulation classification framework (JDM) where a detector predicts frequency centers $f_c$ and bandwidths $B$ to generate proposals that guide the AMC module. The approach demonstrates improved performance over traditional pipelines under varied SNRs, Doppler, clock offsets, and channel models (Rayleigh/Rician), and highlights the practical value of modeling detection before AMC in cognitive radio. Key contributions include the CRML23 dataset, a proposal-driven two-module architecture, and open-source code/dataset to enable reproducible evaluation in realistic spectrum sensing tasks.

Abstract

Signal detection and modulation classification are two crucial tasks in various wireless communication systems. Different from prior works that investigate them independently, this paper studies the joint signal detection and automatic modulation classification (AMC) by considering a realistic and complex scenario, in which multiple signals with different modulation schemes coexist at different carrier frequencies. We first generate a coexisting RADIOML dataset (CRML23) to facilitate the joint design. Different from the publicly available AMC dataset ignoring the signal detection step and containing only one signal, our synthetic dataset covers the more realistic multiple-signal coexisting scenario. Then, we present a joint framework for detection and classification (JDM) for such a multiple-signal coexisting environment, which consists of two modules for signal detection and AMC, respectively. In particular, these two modules are interconnected using a designated data structure called "proposal". Finally, we conduct extensive simulations over the newly developed dataset, which demonstrate the effectiveness of our designs. Our code and dataset are now available as open-source (https://github.com/Singingkettle/ChangShuoRadioData).

Figures (13)

• Figure 1: A typical example of complex signal environment.
• Figure 2: Subfigures (a) and (b) illustrate the total signal count per entry and the distribution of different modulation types within CRML23, respectively. Subfigures (c) and (d) depict the distribution of signal bandwidth and frequency, respectively.
• Figure 3: Joint framework for signal detection and automatic modulation classification (JDM).
• Figure 4: The structure details of the signal detection module.
• Figure 5: The structure of neural network in modulation classification module.