Abstract:In recent years, radar emitter signal recognition has enjoyed a wide range of applications in electronic support measure systems and communication security. More and more deep learning algorithms have been used to improve the recognition accuracy of radar emitter signals. However, complex deep learning algorithms and data preprocessing operations have a huge demand for computing power, which cannot meet the requirements of low power consumption and high real-time processing scenarios. Therefore, many research works have remained in the experimental stage and cannot be actually implemented. To tackle this problem, this paper proposes a resource reuse computing acceleration platform based on field programmable gate arrays (FPGA), and implements a one-dimensional (1D) convolutional neural network (CNN) and long short-term memory (LSTM) neural network (NN) model for radar emitter signal recognition, directly targeting the intermediate frequency (IF) data of radar emitter signal for classification and recognition. The implementation of the 1D-CNN-LSTM neural network on FPGA is realized by multiplexing the same systolic array to accomplish the parallel acceleration of 1D convolution and matrix vector multiplication operations. We implemented our network on Xilinx XCKU040 to evaluate the effectiveness of our proposed solution. Our experiments show that the system can achieve 7.34 giga operations per second (GOPS) data throughput with only 5.022 W power consumption when the radar emitter signal recognition rate is 96.53%, which greatly improves the energy efficiency ratio and real-time performance of the radar emitter recognition system.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to achieve efficient identification of radar transmitted signals in low - power consumption and high - real - time processing scenarios. Specifically, traditional radar transmitted signal identification methods are based on signal parameters (such as carrier frequency, pulse width, signal amplitude, direction of arrival (DOA) and time of arrival of the signal (TOA)), but these methods cannot meet the identification requirements in the modern complex electromagnetic environment. In addition, radar transmitted signals usually need to be intentionally or unintentionally modulated before being radiated into space, which increases the complexity of identification. In recent years, algorithms based on deep learning have been widely used to improve the identification rate of radar transmitted signals because of their powerful feature extraction capabilities. However, these complex deep - learning algorithms and data pre - processing operations have a high demand for computing power, which is difficult to meet in low - power consumption and high - real - time processing scenarios. Therefore, many research works still remain at the experimental stage and cannot be practically applied. To solve this problem, this paper proposes a resource - reuse computing acceleration platform based on field - programmable gate array (FPGA) and implements one - dimensional convolutional neural network (1D - CNN) and long - short - term memory (LSTM) neural network models for radar transmitted signal identification. This model directly classifies and identifies the intermediate - frequency (IF) data of radar transmitted signals, and realizes the parallel acceleration of 1D convolution and matrix - vector multiplication operations by reusing the same systolic array, thereby improving the energy - efficiency ratio and real - time performance of the system while ensuring the identification accuracy. ### Main contributions 1. **Constructed the 1D - CNN - LSTM model**: This model can directly process IF data, ensure high identification accuracy, and has a simple structure, which is convenient for FPGA hardware implementation. Compared with a single CNN, this model has no unified requirement for the length of the input signal and is more suitable for processing pulsed radar transmitted signals. 2. **Designed a resource - reuse computing acceleration platform based on FPGA**: By reusing the same systolic array to realize the parallel acceleration of 1D convolution and matrix multiplication operations, the processing delay is reduced, and at the same time, the utilization rate of FPGA computing resources is greatly improved. 3. **Developed a special instruction set**: According to the different operation characteristics of CNN and LSTM, a special instruction set for the FPGA acceleration platform has been developed, which can be quickly redeployed by adjusting the instructions to adapt to the changes in the NN model structure or algorithm iteration. 4. **Experimental verification**: The 1D - CNN - LSTM radar transmitted signal identification system has been implemented on the Xilinx XCKU040 FPGA development board. The experimental results show that the system achieves a data throughput rate of 7.34 GOPS with a power consumption of only 5.022 W and an identification rate of 97.53%, which is suitable for radar transmitted signal identification scenarios with low - power consumption requirements while ensuring high - computing performance. ### Key technologies of the solution - **1D - CNN - LSTM model**: It combines the advantages of 1D - CNN in local feature extraction and the advantages of LSTM in time - series modeling, and can adapt to randomly changing input signal lengths. - **Systolic array structure**: A 32×64 systolic array is designed to support the parallel calculation of 1D single - channel convolution and matrix - vector multiplication. - **Fixed - point number representation**: All data participating in the operation are 16 - bit fixed - point numbers, including 1 - bit sign bit, 4 - bit integer bit and 11 - bit decimal bit, in order to reduce the consumption of computing resources. - **Cache structure**: A multi - channel 1D convolution cache structure is designed. By reading the result of the previous channel, adding it to the result of the current channel, and then storing it back to the original address, the efficient calculation of multi - channel convolution is realized. Through these technologies, this paper has successfully solved the problem of achieving efficient identification of radar transmitted signals in low - power consumption and high - real - time processing scenarios.

Efficient FPGA Implementation of Convolutional Neural Networks and Long Short-Term Memory for Radar Emitter Signal Recognition

FPGA Implementation of High-Speed Data Acquisition System for High-Resolution Millimeter Wave Radar.

HLS-Based FPGA Implementation of Multi-Label Learning for Radar Signal Recognition

Automatic Deployment of Convolutional Neural Networks on FPGA for Spaceborne Remote Sensing Application

FPGA Implementation of Efficient CFAR Algorithm for Radar Systems

A Resource Efficient CNN Accelerator for Sensor Signal Processing Based on FPGA.

Design and Implementation of a FPGA-based Accelerator for Convolutional Neural Networks

An Efficient Algorithm-Hardware Co-Design for Radar-Based Fall Detection With Multi-Branch Convolutions

A Low-Cost Fully Integer-Based CNN Accelerator on FPGA for Real-Time Traffic Sign Recognition

Design and Implementation of Convolutional Neural Network Image Recognition Acceleration System Based on FPGA

FPGA-based Neural Network Accelerator for Millimeter-Wave Radio-over-Fiber Systems

Specific Radar Recognition Based on Characteristics of Emitted Radio Waveforms Using Convolutional Neural Networks

An FPGA-Based Energy-Efficient Reconfigurable Convolutional Neural Network Accelerator for Object Recognition Applications

A Memory-Optimized and Energy-Efficient CNN Acceleration Architecture Based on FPGA.

FPGA Implementation of an Efficient FFT Processor for FMCW Radar Signal Processing

NeuroREC: A 28-Nm Efficient Neuromorphic Processor for Radar Emitter Classification

An Efficient FPGA-Based Implementation for Quantized Remote Sensing Image Scene Classification Network

Efficient Neural Networks on the Edge with FPGAs by Optimizing an Adaptive Activation Function

A Grain-Adaptive Computing Structure for FPGA CNN Acceleration.

Radar Emitter Signal Recognition Based on One-Dimensional Convolutional Neural Network with Attention Mechanism

FPGA-based Accelerator for Convolutional Neural Network