Many techniques for recognizing and identifying disturbed signals waveforms are primarily based on visual inspection. This paper proposes a wavelet packet decomposition based technique to perform a feature extraction from the disturbed signals in order to identify the possible causes of the disturbance. On the basis of definition groups of Electromagnetic Interference, the interested information about the EMI source by Bayesian discrimination method was obtained. Especially, after applied Parseval's theorem energy rule, it was greatly reduced the volume of computation. The computation example had been used to illustrate its validity in the vehicle Electromagnetic Compatibility design.