The detection of the engine rotational speed in revolutions per minute (RPM) is of great importance to estimate the speed of boats. This value can be obtained from the fundamental frequency component of acquired sonar signals. However, detection can often be seriously affected by noise and distortion introduced by the underwater environment. Several methods have been proposed for fundamental component detection, but they do not specifically take advantage of the passive sonar signal characteristics to improve the performance of the process. In this context, the proposed algorithm uses DEMON processing applied to wavelets packets subbands to exploit the characterization of the sonar signal in the time and frequency domains. The algorithm involves signal segmentation, wavelet packet decomposition, subband envelope cross-correlation and fundamental component detection from the power spectrum. The method was applied in passive sonar signals acquired in navigation and also obtained by simulation. The performance of the proposed algorithm was evaluated with signals of different SNR values that were also corrupted by a simulated multipath underwater channel. The signals were evaluated by both the experienced sonar operators and the proposed algorithm. The results obtained were very satisfactory for RPM detection and are detailed at the end of this document.