Abstract: Transient electromagnetic signals have a short duration, and there are many redundant data such as noise and interference when carrying out data acquisition and signal processing. The values of kurtosis have significant advantages in characterizing pulse signals such as transients and shocks. Based on kurtosis and sliding window, the detection and extraction method of transient electromagnetic signals under noise and interference has been studied in this paper. The analysis results show that the kurtosis eigenvalues between the transient electromagnetic signal and the noise interference signal have high distinguish ability, good stability and little influence by the changes of signal strength and other parameters. It can be used to detect and extract single or multiple transient electromagnetic signals. This method can effectively reduce the proportion of invalid data such as noise and interference signals, it has less calculation, so it is simple and fast. Simulation and experimental results also verify the feasibility of this method.