Automatic recognition of modulation scheme in blind environment plays a key role in many communication applications. A hierarchical and local density (HLD) approach is proposed to classify eight modulation schemes in a two stage process. In the first stage, the domain of modulation schemes (FSK, ASK, PSK, and QAM) is identified. FSK is identified based on feature extracted through complex envelope of downconverted signal. ASK scheme is identified using linear regression error. PSK and QAM modulation schemes are recognized based on the ratio of sixth and fourth order cumulant. In the latter stage, the order of modulation (ASK, PSK, and QAM) is classified through its respective ideal constellation points. HLD can correctly identify 2ASK, 4ASK, and QPSK modulation schemes in AWGN channel above 8 dB SNR and the other modulation schemes (8ASK, 8PSK, 16QAM, and 64QAM) above 16dB SNR. HLD is implemented in NI labVIEW and validated on the signals generated through PXIe-5673 and received using NI PXIe-5661. The proposed HLD classifier does not require any training to set thresholds as compared to more complex SVM, KNN, and Naive Bayes Classifier based techniques and shows an improved accuracy.

在盲环境中自动识别调制方案在许多通信应用中起着关键作用。提出了一种分层和局部密度（HLD）方法，以在两个阶段的过程中对八个调制方案进行分类。在第一阶段，识别调制方案（FSK，ASK，PSK和QAM）的域。基于通过下变频信号的复杂包络提取的特征来识别FSK。使用线性回归误差识别ASK方案。基于六阶和四阶累积量之比来识别PSK和QAM调制方案。在后面的阶段中，调制顺序（ASK，PSK和QAM）通过其各自的理想星座图点进行分类。HLD可以正确识别高于8 dB SNR的AWGN信道中的2ASK，4ASK和QPSK调制方案，以及其他调制方案（8ASK，8PSK，SNR高于16dB的16QAM和64QAM）。HLD在NI labVIEW中实现，并通过PXIe-5673生成并使用NI PXIe-5661接收的信号进行了验证。与更复杂的基于SVM，KNN和朴素贝叶斯分类器的技术相比，提出的HLD分类器不需要任何培训来设置阈值，并且显示出更高的准确性。