Abstract: In the routine infrared inspection signal processing, pulse phase thermography doesn't take into account the measure noise, the experimental data isn't made good use of by the simple Fourier transform, where more information needs to be dig. The non-Gaussianity analysis was introduced to the routine signal processing, the third-order cumulant and bispectrum estimation were used for extracting the characteristic information from the detection signal when damage occurred, therefore the criterion of distinguishing sound area from defective area was obtained. The definitions of third-order cumulants and bispectrum were achieved for the acquired signal, in addition, the advantage using magnitude and phase of slice spectrum was analyzed and the effects were compared in actual detection. The proposed method has the advantages of suppressing Gaussian and symmetry distribution non-Gaussian noise efficiently, small calculation amount, fast computation speed. Moreover, it compensates the demerit which magnitude of spectrum is hard to use in the traditional pulsed phase thermography.