Extraction of Small Resonance Signals in Strong Same-Frequency Background

In practical engineering applications, attenuated sinusoidal signals are very common, but in most practical scenarios, such signals will be submerged in a strong background. When the frequency of the background signal is the same as the signal to be measured, this signal extraction of the difficulty of the process will increase sharply. This paper compares the effects of short-time Fourier transform, continuous wavelet transforms, and shallow neural networks on this practical physical problem. In the end, the results of the shallow neural network are relatively good, achieving a demodulation error of one thousandth, which is an effective solution to this physical problem.