Related Experiment Video
Updated: Aug 29, 2025
06:00
Electroantennography-based Bio-hybrid Odor-detecting Drone using Silkmoth Antennae for Odor Source Localization
Published on: August 27, 2021
5.4K
SAS-SEINet: A SNR-Aware Adaptive Scalable SEI Neural Network Accelerator Using Algorithm-Hardware Co-Design for
Jiayan Gan1,2, Ang Hu1, Ziyi Kang1
1School of Information and Communication Engineering, University of Electronic Science and Technology of China, Chengdu 611731, China.
Sensors (Basel, Switzerland)
|September 9, 2022
Summary
This study introduces a power-efficient hardware accelerator for specific emitter identification (SEI) in unmanned aerial vehicle (UAV) surveillance. The design achieves high accuracy by adapting deep convolution neural networks (DCNNs) to varying signal conditions.
Area of Science:
- Electrical Engineering
- Computer Science
- Aerospace Engineering
Background:
- Radio Frequency (RF)-based surveillance using unmanned aerial vehicles (UAVs) is crucial for air control.
- Specific Emitter Identification (SEI) technology analyzes RF signals for identifying controllers.
- Existing SEI algorithms lack dedicated hardware implementations.
Purpose of the Study:
- To propose a high-accuracy and power-efficient hardware accelerator for SEI in UAV surveillance.
- To develop an algorithm-hardware co-design approach for enhanced SEI performance.
- To address the need for specialized hardware in deep convolution neural network (DCNN)-based SEI.
Main Methods:
- Developed a scalable SEI neural network with SNR-aware adaptive precision computation.
- Integrated Short-Time Fourier Transform (STFT) with DCNN for feature extraction.
- Designed a hardware accelerator featuring SNR sensing, denoising, and specialized DCNN engines with hybrid precision.
Main Results:
- Achieved a maximum accuracy of 99.3% and an F1 score of 99.3% on a public UAV dataset.
- Demonstrated power efficiency of 40.12 GOPS/W (INT16) and 96.52 GOPS/W (binary precision).
- Validated the design's effectiveness on a Field-Programmable Gate Array (FPGA).
Conclusions:
- The proposed algorithm-hardware co-design significantly improves SEI accuracy and power efficiency for UAV surveillance.
- SNR-aware adaptive precision computation enables robust performance across varying signal-to-noise ratios (SNRs).
- The specialized hardware accelerator provides a practical solution for real-time SEI tasks in UAV applications.
