In order to detect the weak signal from the strong noise, the excellent time-frequency localization analysis properties of wavelet packet transform is analyzed according to the different Lipschitz exponent the frequency spectrum of signal and noise are analyzed by introducing ∞ - norm. Then, the optimal subspace entropy and the position parameters of the intact bintree in optimal subspace are regarded as the characteristic quantities, the low SNR radar signal detection is realized by using the method of floating threshold denoising. Lastly, the low SNR radar signal detection based on wavelet packet transform and wavelet transform is simulated through numerical calculation in computer, at the same time, the conclusion that the wavelet packet transform is better than the wavelet transform in the low SNR signal detection is drawn.