No abstract

The paper introduces a novel method to detect the initial time of a weak sinusoidal signal embedded in non-stationary noise and studies the performance of the method through numerical simulations. Furthermore, a proposed nonlinear function combined with time-frequency analysis can be used to effectively improve the signal-to-noise (SNR). Simulation results show that the method can rapidly detect the initial time of the weak analytical signal with a SNR lower than −54 dB and a time-domain resolution of only 1 second.