摘要
本文提出了一种从观察序列的广义自回归(GAR)模型参数提取待识别信号的伪瞬时中心频率和伪瞬时3dB带宽特征,并利用神经网络分类器的数字调制识别新方法。这种方法充分利用了GAR模型良好的抗噪声能力和神经网络优异的模式分类能力,能有效地改善低SNR条件下的调制识别性能。计算机模拟结果证实了该方法具有很高的识别率和良好的稳健性。
A new digital modulation recognition method is proposed, which is based on the features extracted from generalized autoregressive (GAR) model parameters of the obsevered sequence and the use of neural network classifier. Because of the better noise suppression capacity of the GAR model and the powerful pattern classification capacity of the neural network classifier, the new method can significantly improve the recognition performance in lower SNR. Computer simulations also show that the new method possesses higher recognition ratio and better robustness.
关键词
调制识别
广义自回归模型
神经网络分类器
Modulation recognition, GAR model, Feature extraction, Neural network classifier
