| Abstrakt: |
In response to the challenges posed by low Signal-to-Noise Ratio(SNR) and the variability in symbol oversampling rates after baseband transformation, a modulation recognition algorithm leveraging high-order cumulant features and P-LSTM networks is proposed, which utilizes second-order, fourth-order and sixth-order cumulant of root-raised cosine shaped filtering signals,extracting the feature parameters in the form of their absolute values, and a two-layer independent parallel Long Short-Term Memory (P-LSTM) network to recognize five modulation signals: BPSK, QPSK, 8PSK, 16QAM and 32QAM. Simulation experiments demonstrate that the proposed method achieves a recognition rate of 99.3% at SNR≥5dB and over 99% recognition from 5 times oversampling to 20 times oversampling. [ABSTRACT FROM AUTHOR] |
