Demonstrates a CNN-based framework that improves classification accuracy in wireless communication under various noise conditions, suggesting better real-world applications.
Key Points
To develop a robust automatic modulation classification framework using deep learning methods that perform well under low signal-to-noise ratio conditions.
Developed a customized dataset using MATLAB for control over modulation types and channel conditions.
Employed STFT-based spectrogram representations to extract features from modulated signals.
Trained a CNN model to perform automatic modulation classification on the dataset.
Achieved a classification accuracy of 99.77% at 15 dB.
Maintained robust performance across an SNR range of -5 to 15 dB.
Improved generalization capability and reduced reliance on pre-existing datasets, supporting eleven modulation schemes.