High speed train noise level is an important factor with respect to passenger comfort and life quality of residents along the railway. How to attenuate the noise level is an important research direction that train designers care about. Studies show that train interior noise is consist of trans- mission noise and structural noise. Separating these two kinds of components from their overall ob- servations will provide further guide to high speed train noise reduction design. The research is based on the real-world high speed train noise data. First, data are analyzed by different digital sig- nal processing methods and some basic properties of the train noise signal are concluded. Then, by modeling and analyzing the noise transmission circumstances when train is stationary and moving, train body frequency response is used to measure the noise insulation quality, and an algorithm is proposed to calculate the frequency response conveniently. At last, train body frequency response is calculated according to the real-world train noise data via the proposed algorithm, and the result is used to separate transmission noise and structural noise from the recorded noise data.