The normal operation of marine diesel engines ensures the scheduled completion and efficiency of a trip. Any failures may result in significant economic losses and severe accidents. It is therefore crucial to monitor the engine conditions in a reliable and timely manner in order to prevent the malfunctions of the plants. This work describes and evaluates the development and application of an intelligent diagnostic technique based on the integration of the empirical mode decomposition (EMD), kernel independent component analysis (KICA), Wigner bispectrum and support vector machine (SVM). It is an extension of the previous work on the fault detection for a diesel engine using the instantaneous angular speed (IAS). In this study, in order to solve the underdetermined blind source separation (BSS) problem the combination of EMD and KICA is firstly presented to estimate IAS signals from a single-channel IAS sensor. The KICA is also applied to select distinguished features extracted by Wigner bispectrum. The SVM is then employed for the multi-class recognition of the marine diesel engine faults in an intelligent way. Numerical simulations using a 6-cylinder engine model and real IAS data measured on the ship named "Hangjun 20" are used to evaluate the proposed method. Both the numerical and experimental diagnostic results have shown high efficiency of the proposed diagnostic method. Distinct fault features of the IAS signals have been extracted by the EMD-KICA and Wigner bispectrum, and the fault detection rate of the SVM is beyond 94.0%. Thus, the proposed method is feasible and available for the fault diagnosis of marine diesel engines.