TY - JOUR
T1 - Blind Modulation Recognition of Overlapping Signals via Compressive Sensing-Based Higher-Order Cyclic Cumulant Recovery
AU - Liu, Minrui
AU - Zhang, Zehui
AU - Xu, Haixiang
AU - Hou, Chaoqun
AU - Kang, Junjie
AU - Ye, Neng
N1 - Publisher Copyright:
© 2026 IEEE.
PY - 2026
Y1 - 2026
N2 - In satellite communication scenarios, multiple signals may coexist in adjacent bands or shared channels because of limited spectrum resources and multi-user access. Their spectral overlap enlarges the effective received bandwidth and couples the modulation features of different components, which makes blind modulation recognition difficult and also increases the burden on wideband sampling. This paper proposes a blind modulation recognition method based on compressive sensing and higherorder cyclic cumulants. The proposed framework acquires compressed observations through pseudo-random sub-Nyquist sampling and directly recovers high-order cyclic cumulant spectra by exploiting their sparsity in the cycle-frequency domain. The additive property of higher-order cyclic cumulants further allows the modulation-related structural peaks of different components to remain distinguishable in the recovered spectra, enabling blind recognition through structural feature matching. Simulation results show that, with sufficient temporal accumulation, the proposed method achieves more than 85% recognition accuracy at 3 dB SNR when the compressed observations are reduced to one quarter of the original sampled data.
AB - In satellite communication scenarios, multiple signals may coexist in adjacent bands or shared channels because of limited spectrum resources and multi-user access. Their spectral overlap enlarges the effective received bandwidth and couples the modulation features of different components, which makes blind modulation recognition difficult and also increases the burden on wideband sampling. This paper proposes a blind modulation recognition method based on compressive sensing and higherorder cyclic cumulants. The proposed framework acquires compressed observations through pseudo-random sub-Nyquist sampling and directly recovers high-order cyclic cumulant spectra by exploiting their sparsity in the cycle-frequency domain. The additive property of higher-order cyclic cumulants further allows the modulation-related structural peaks of different components to remain distinguishable in the recovered spectra, enabling blind recognition through structural feature matching. Simulation results show that, with sufficient temporal accumulation, the proposed method achieves more than 85% recognition accuracy at 3 dB SNR when the compressed observations are reduced to one quarter of the original sampled data.
KW - Overlapping signals
KW - automatic modulation recognition
KW - compressive sensing
KW - higher-order cyclic cumulants
KW - satellite communications
UR - https://www.scopus.com/pages/publications/105052975412
U2 - 10.1109/ICCC70295.2026.11680305
DO - 10.1109/ICCC70295.2026.11680305
M3 - Conference article
AN - SCOPUS:105052975412
SN - 2377-8644
JO - IEEE International Conference on Communications in China Workshops, ICCC
JF - IEEE International Conference on Communications in China Workshops, ICCC
IS - 2026
T2 - 2026 IEEE/CIC International Conference on Communications in China, ICCC 2026
Y2 - 7 August 2026 through 9 August 2026
ER -