On-Chip Acceleration of RF Signal Modulation Classification With Short-Time Fourier Transform and Convolutional Neural Network

Abstract

Automatic Modulation Classification (AMC) is a technique used in wireless communication systems to identify the modulation type of received signals at the receiver. Improving spectrum utilization efficiency is essential for AMC and is carried out by fundamental signal processing methods within the physical layer. Convolutional Neural Networks (CNN) based deep-learning models have recently been employed in AMC systems, demonstrating superior performance. However, the large size of CNN models, floating-point weights, and activations make deploying such systems with limited hardware resources quite complex. In this paper, we propose a hybrid Radio Frequency-based Machine Learning (RFML) model that combines Short-time Fourier Transform and Convolutional Neural Network (STFT-CNN) for AMC. Simulations on RadioML2016.10a demonstrate an average recognition accuracy of 79% for a Signal-to-Noise Ratio of 0dB or higher when using 16-bit fixed-point operations. An on-chip accelerator for the STFT-CNN, designed and synthesized in 28nm CMOS, offers 19 times lower power consumption, 34 times smaller area, 1.3 times higher bandwidth, and 50 times less memory than a time-domain CNN accelerator utilizing 32-bit floating-point operations.

Keywords

• Radio frequency machine learning
• convolutional neural network (CNN)
• short time Fourier transform (STFT)
• hardware accelerator