Cross-hole sonic logging has been widely used in the detection of the integrity of bored pile, which use the acoustic parameters(velocity, wave amplitude, product of the slope of two adjacent lines on the time-depth curve and acoustic time difference(PSD) and the extent of waveform distortion to determine the classification of pile integrity, but the extent of waveform distortion can not be expressed quantitatively. The waveform distortion coefficient and the critical value of distortion coefficient is calculated by the point energy statistics method in frequency domain, using the spectrum analysis and wavelet packet analysis method to explore the features of normal waveform and distortion waveform (the measuring point of pile with bottom sediment, low intensity and non dense pile, broken pile), and the feature vector is structured via the frequency band energy of 3 layer wavelet packet decomposition signal. The results show that the spectrum function for the measuring point on the pile with bottom sediment is stepped, the spectrum energy for the measuring point on broken pile is mainly distributed in the low frequency part; and the eigenvector of the normal measuring points is the largest; the broken pile is the smallest; the non-dense measuring points with low intensity have 1 upper convex point.