The Middle Miocene Minghuazhen Formation of the Bohai Bay Basin is dominated by fluvial channels and shallow-lacustrine deltaic systems. These depositional facies, along with complex postdepositional faulting, make it difficult to detect fluvial channels. As a useful seismic attribute to solve this problem in the Bohai Bay Basin, spectral decomposition of 3D seismic data can provide significant information to understand the subsurface fluvial channels. The analytic wavelet transform (AWT) is a promising approach for implementing spectral decomposition to provide a detailed time-frequency representation. In particular, by varying two parameters (beta and gamma) controlling the wavelet forms, the generalized Morse wavelets (GMWs) can be given a broad range of characteristics while remaining exactly analytic. To detect fluvial channels with thickness around and lower than the tuning thickness, we have proposed a suitable (beta = 1 and gamma = 3) pair of parameters for GMWs because such an exactly analytic Morse wavelet substantially outperforms the approximately analytic Morlet wavelet for high time localization. We have applied the AWT with a GMW (beta = 1 and gamma = 3) to 3D seismic data in the Lower Minghuazhen Formation, Bohai Bay Basin, China. The stratal slicing of spectral decomposition volumes at depositional cycle 2 of the Lower Minghuazhen Formation shows its potential for channel interpretation and the optimization of well patterns.