Chord measurement is a widely-used method of inspecting track irregularity at present. Considering the singular characteristics of rail track surface defects, a new method is proposed to identify these defects on the basis of multi-chord theory by analyzing their responses to mid-chord model. Then, the test data are separated by singular value decomposition, since the surface damage is enhanced when the track irregularity amplitude gain is reduced by proper combination of chords. Thus, the rail track surface defects characteristics and their positions are extracted through the sliding kurtosis sequence. Theoretical simulation and test results show that this approach is able to get rail track surface defects information which is hidden in chord measured value. The proposed method is suitable for engineering application.