TY  - JOUR
T1  - Enhanced Automatic Modulation Recognition Performance Using ResNet and GAF
AU  - Lee, Sangho 
AU  - Lim, Wansu 
JO  - The Journal of Korean Institute of Communications and Information Sciences
PY  - 2025
DA  - 2025/1/1
DO  - 10.7840/kics.2025.50.4.595
KW  - Automatic Modulation Recognition(AMR)
KW  - Gramian Angular Field(GAF)
KW  - Residual Neural Network(ResNet)
AB  - Automatic Modulation Recognition (AMR) is a crucial technology for enhancing the reliability and efficiency of wireless communication. Recent AMR research has focused on improving performance by utilizing deep learning, with various models being proposed that use IQ constellations as training data. However, since constellations are tools for visualizing the signal state at a specific time, they fail to capture the temporal characteristics of the signal, which can lead to performance degradation. To address this issue, this paper proposes a novel visualization method called GAF that incorporates the temporal features of time-series data, along with the application of PCA for channel reduction and noise removal. The proposed GAF method effectively captures the temporal variations and patterns in the signal, providing more diverse information than traditional visualization techniques. Furthermore, applying this approach to a ResNet model resulted in 5% to 15% higher accuracy in the SNR range of -10dB to 5dB compared to other visualization techniques such as IQ constellations, GCC algorithms, and the traditional GAF method used in AMR.