Table 3. Comparison of average recognition accuracy of different SEI algorithms under different input features.

SEI algorithm	Sampling signal form	Number of sample points utilized	Input feature form	Feature size	Average accuracy rate
Proposed algorithm	4000 points of baseband complex sequence signal sampling points	2000	Time domain I/Q sequences Time domain A/P sequence Frequency domain A/P sequences	6×2000	0.737
Vector map algorithm		4000	Grayscale vector image	300×300×3	0.570
Variational modal decomposition algorithm		4000	Second eigenmode function signal sequence	1×4000	0.222
Multi-domain feature fusion algorithm		4000	Time domain I/Q sequences Power Spectrum Sequence I-way integral bispectral sequence Q-way integral bispectral sequence	5×1280	0.429
Multi-projection feature fusion algorithm		4000	Two-dimensional projection of wavelet eigencoefficient matrix Two-dimensional projection of the bispectral eigencoefficient matrix Two-dimensional projection of the Hebert sign coefficient matrix	224×224×3 224×224×3 224×224×3	0.326
Proposed algorithm		4000	Second eigenmode function signal sequence	1×4000	0.267
Proposed algorithm		4000	Time domain I/Q sequences Power Spectrum Sequence I-way integral bispectral sequence Q-way integral bispectral sequence	5×1280	0.680