RF Signal-Based UAV Detection and Mode Classification: A Joint Feature Engineering Generator and Multi-Channel Deep Neural Network Approach

. 2021 Dec 14;23(12):1678. doi: 10.3390/e23121678

Algorithm 1 Feature Engineering Generator Algorithm.

Require:

The original low-frequency time domain component L.
The original high-frequency time domain component H.
The number of samples in each data segment M.
The number of categories of data from different types of UAVs with flight modes N.
The points of moving average filter for the low-frequency and high-frequency components $n_{l}$ and $n_{h}$ , respectively.

Ensure:

The Feature Engineering Generator preprocessed frequency domain data D.

1:
forn in N do
2:
Extract the time domain low-frequency component $L_{n}$ and high-frequency component $H_{n}$ of category n.
3:
Resegment $L_{n}$ , $H_{n}$ into new segments with M samples per segment $S_{L}$ , $S_{H}$ , respectively.
4:
for l in $S_{L}$ do
5:
Fourier transform l.
6:
$n_{l}$ -point moving average filter l.
7:
end for l
8:
for h in $S_{H}$ do
9:
Fourier transform h.
10:
$n_{h}$ -point moving average filter h.
11:
end for h
12:
$S_{L} = \frac{S_{L}}{max (S_{L})}; S_{H} = \frac{S_{H}}{max (S_{H})}$ .
13:
$S = (S_{L}; S_{H})$ .
14:
$D_{n} = (S^{2}; n)$ .
15:
end forn
16:
$D = (D_{1}, D_{2}, \dots, D_{n})$ .