The signal acquisition of electric vehicles is complex and variable. This study applied the traditional empirical mode decomposition (EMD) method to analysis the multi-frequency signals of electric vehicle fault signals. It was found that such signals mixed in colored noise is harmful for signal decomposition. In a classical bistable system, under the synergistic effect of adaptive stochastic resonance and EMD, the noise signal decomposition fails. Then, this article proposes a method that combines adaptive SR and EMD in periodic potential systems to achieve decomposition. At the same time, the Random Particle Swarm Optimization (RPSO) method was applied to determine optimal signal-to-noise ratio under condition of it reaches the peak value. According to the simulation results, it can be inferred that this method is effective and valuable, and is beneficial for signal processing in such field.