Dealing with low probability of intercept (LPI) radar signals in very low signal-to-noise ratio (SNR) stages requires reconnaissance systems to employ an effectively practical method for detecting and fully characterising the received signals. Following the aim of LPI signal detection, in this study, a novel method based on concentrating the energy of the signal in one particular row of the time–frequency domain matrix is proposed. It is shown that the proposed method enables to detect and estimate the parameters of multiple low-power linear-frequency-modulated signals. A comparison has been made between the performance of the matched filter and the proposed method, and it is shown the proposed method has approximately the same performance and low computational complexity as matched filters. This method is also able to detect the poly-phase signals such as Frank-coded signals and also non-linear-frequency-modulated signals, as well. To show the effectiveness of the method, extensive simulations are carried out with different LPI radar waveforms corrupted with additive white Gaussian noise of SNR down to −25 dB.