This repository contains information, code from the manuscript [Deep-Learning-Based Multipath Time of Arrival Estimation using Sparse Kernel Representation], which is based on Pytorch. And the code contains the core content of the proposed algorithm.

data.re.py lists ont-hot, correlation and gaussian coding for off-grid time delay representation.

The network of proposed method as shown in modules.py.

AS shown in data.data.py, we generate the simulation chirp signal with parameters as:

train.py provides the code for training a model from scratch.

The network include 7 TE module, and a module include 2 CNN, Relu and Batchnormalization, input is added to the output of second BN as residual block. And the SE is added to the residual block as TE module.

The parameter as shown in modules.py.