A time‐frequency method for detecting VHF underdense meteor signals

Underdense meteor echoes observed using very high frequency (VHF) radar can be accurately modeled as a single complex damped sinusoid in additive white Gaussian noise. The normalized damping coefficient is expected to be between 0 and 0.3 for a VHF meteor return based on the modeled ambipolar diffusion rates near 90 km and a radar operating between 30 and 50 MHz. Current meteor echo detection routines operate either in the frequency domain, where it is difficult to detect highly damped signals, or in the time domain, where it is difficult to detect narrow band signals. An added difficulty is that typical approaches require a priori knowledge of the noise variance which can bias the performance of these estimators. In this paper, a time‐frequency waveform detector is proposed to address these problems. By normalizing the signal power and power spectral density in the time and frequency domains, respectively, a detector that is invariant to the noise variance can be implemented. The characteristics of a damped sinusoid in the time and frequency domains are exploited to construct the detector. The threshold for the proposed detector is only a function of of the time series length, and is stable with respect to a range of damping coefficients and noise levels. The time‐frequency waveform detector exhibits superior performance to the conventional energy or power detectors when the signal of interest is highly damped or the noise variance is unknown. A derivation of the time‐frequency waveform detector, comparison with the energy and power detectors and numerical results demonstrating the effectiveness of this detector are presented.