As more and more power electronic equipments are incorporated into the grid, the new power system is becoming increasingly unstable. This is usually due to the impedance mismatch between different power electronic converters. Therefore, accurate impedance measurement technology plays an important role in ensuring the stability of power system. Multi-sinusoidal signal has become a common impedance measurement method because of its accuracy and rapidity. The traditional multi-sinusoidal signal is selected by equidistant multiple sinusoidal signals. In the absence of guiding theory, the selection of frequency interval always relies on experience accumulation, resulting in low accuracy, low reliability and large manual intervention. In order to solve the above problems, a Dynamic Mean Thresholding (DMT) method for adaptive frequency selection of multiple sinusoidal signals is proposed. The DMT method calculates the average error of all impedance measurement points, and then selects the points with greater than the average error for the next iteration. Compared with traditional methods, DMT method has higher measurement accuracy and less manual intervention. Finally, the proposed impedance measurement scheme is verified by HIL experiment platform.