Nowadays, the acoustic detection is widely used for defect diagnosis of gas insulated substations (GIS) in normal operation and factory tests. In this paper in order to develop a data analyzer for acoustic detection system to make an assistant diagnosis, the characteristic of acoustic signals generated by different artificial defects such as protrusions, floating shield, void in spacer and bouncing particles are investigated. Some meaningful parameters behind the detected acoustic signals are extracted and discussed, which are used to distinguish background noise, partial discharge (PD) phenomena or bouncing particles. Based on those works, a comprehensive identification method realized by processing the acoustic pulse sequences qi, ( Δ ti, qi) and (ti, qi) is introduced, which gives a recognition result with noise, PD type or bouncing particles. For the sequence (ti, qi), the backpropagation artificial neural network optimized by genetic algorithm (GA-BPANN) is used as a classifier based on the fingerprint consisting of 24 operators, which are derivate from typical 2D histograms of phase-resolved partial discharge (PRPD). And with considering the trigger source may having a phase difference from the working voltage, identification with phase compensation (IPC) is used as a try to deal with the challenge. Experimental results show that the comprehensive identification method is practical and effective.