The FM (Frequency Modulated) LPI (Low Probability of Intercept) radars, unlike conventional radars, transmit a long or a continuous signal, with low power, using causal modulation. These LPI radar characteristics make it virtually invisible to conventional RWRs (Radar Warning Receivers) and ESM (Electronic Support Measures) receivers. Many studies in the EW (Electronic Warfare) field are being carried out to deal with this LPI radar advantage. One promising approach is the employment by ESM equipments of time-frequency transform methods to find causality in the spectrum's noise. This paper applies the WPT (Wavelet Packets Transform) to the matter of generating a portrait of the electromagnetic spectrum, aiming the milestone for an automatic classification method. We compare the time-frequency portrait of FM LPI radar signals obtained through the WPT and the well-known Choi-Williams and Fourier transforms. Results obtained from real data show some advantages to the WPT.