Regarding the difficulties of detecting the different degrees of blocking in buried sewers and efficiently extracting effective blockage characteristics， a blocking detection method using acoustical characteristics based on information gain and extreme learning machine is proposed in this paper. Firstly， a sinusoidal sweep signal is used as the excitation signal， and the sound pressure data are collected at the receiving end to measure the acoustic impulse response of the pipe. Then a six-level wavelet decomposition tree is applied to obtain eight decomposed acoustic signal components in different frequency bands. After that， the information gain is introduced to quantitatively characterize the eight components， and the components that contain the most information about the pipe conditions are chosen so that the effective feature information can be kept. The sound pressure level is calculated to form the eigenvectors from the Mel-frequency cepstrum coefficients. Finally， the extreme learning machine is adopted due to its advantages of simple structure and fast learning speed. So， the pipe conditions can be effectively identified， as well as the different degrees of blocking of the pipe through the above methods， that the experimental results verify the effectiveness. Furthermore， it can also eliminate the interference of pipe components such as lateral connection to improve the accuracy of blockage identification， which has both theoretical significance and application value for condition assessment in sewer system.