Abstract: Objective: The operation control system of urban rail transit trains mainly relies on wireless communication, and external interference poses a serious threat to its normal operation. To promptly detect external sources of interference and ensure the safety of wireless communication systems in rail transit, it is necessary to study the adaptive VMD (variational mode decomposition) algorithm used in RFF (radio frequency fingerprinting) identification technology, aiming to improve the accuracy rate of identifying wireless transmitter devices in rail transit environment.Method: After extracting RFF, the adaptive VMD algorithm is applied. Two indicators, namely the correlation coefficients between different modes and the proportion of each mode energy in the total signal energy, are utilized to jointly determine whether mode aliasing or overdecomposition occurs during the decomposition process. Based on this, an iterative approach is employed to select appropriate decomposition mode number and penalty factor, thereby enhancing the accuracy of mode decomposition. In the identification process, the reconstructed signal composed of the different modes is used as the device RFF, and the LSTM (long shortterm memory) is employed for wireless transmitter device classification and identification.Result & Conclusion: Experimental results demonstrate that the reconstructed signal obtained through the adaptive VMD algorithm not only preserves relatively complete fingerprint characteristics compared to the original signal, but also suppresses noise to a certain extent, exhibiting strong RFF extraction capability. In the identification of actual WiFi (wireless fidelity) devices, the identification accuracy of the adaptive VMD algorithm is significantly superior to existing similar programs with less susceptibility to noise, indicating that the algorithm can effectively improve identification accuracy rate of WiFi emission devices.