Modulation Recognition using Wavelet Transform based on AlexNet
Jie Yang, Fan Liu · 2019
In this paper, a modulation recognition method is presented to implement the complex modulation recognition in non-cooperation communication systems by using an AlexNet-based deep learning technique. In the image preprocessing step, we convert the time-domain diagrams of different complex modulated signals to the spectrogram images by using wavelet transform (WT). In the training step, with the feature extraction of AlexNet model to the input spectrogram images, we can attain the training model, which can be used to classify the different modulations of signals. In this work, we select the eight modulated signals (2ASK, 4ASK, 2PSK, 4PSK, 2FSK, 4FSK, 16QAM, 64QAM) for recognition. The simulation result shows that the classification accuracy of eight signals is almost 100%at higher signal-to-noise ratio (SNR). Besides, the experimental result also illustrates the similarity between the spectrogram images of 16QAM, 4PSK and 4ASK. Compared with using initial time-domain diagrams without wavelet transform as input, we confirm that the superiority of the presented method whose classification accuracy is significantly improved.