“基于复杂神经网络的双基地ISAC联合AoA和AoD估计”的matlab实现.zip

# Bistatic_ISAC # This is the code base for our paper "Complex Neural Network based Joint AoA and AoD Estimation for Bistatic ISAC". Authors: Salmane Naoumi, Ahmad Bazzi, Roberto Bomfin, Marwa Chafii - - - - ## Requirements ## Requirements are listed in the file **requirements.txt** ```bash pip install -r requirements.txt ``` ## Simulation Data ## Simulation data are generated using MATLAB and saved as .mat files 2D Estimator and CRB results are also obtained using the MATLAB scripts **CRB3D.m** and **Estim2D_new** The scripts to run in MATLAB to generate the simulation data for training are: - SNRAnalysis.m - ClassifyNumberTargets.m - Compare1TargetPeak.m - Compare2TargetsPeak.m ## Joint Estimation Results ## <p align="center"> <img src="https://github.com/salmane-s9/Bistatic_ISAC/blob/main/images/image_model.png" alt="drawing" width="500"/> </p> ### SNR Analysis ### This script plots the performance of the MLP architecture trained at different SNR values and used for prediction on a range of Test SNRs (Figure. 4) ```bash python snr_analysis.py ``` ### Number of Targets classification ### This script plots the performance of the MLP Classfier architecture for predictiing the number of targets in one peak (Figure. 5) ```bash python classif_ntargets.py ``` ### Joint AoA-AoD predictions ### This script plots the performance comparison of MLP and convolutional networks with the parametric 2D Estimation algorithm for AoA estimation (Figure. 6) ```bash python compare_1target_peak.py ``` This script plots the performance comparison of MLP with the 2D Estimation algorithm for both AoA and AoD angles for settings with two targets per peak (Figure. 7) ```bash python compare_2targets_peak.py ```