The acoustic signals were strongly related to the behavior of the arc column, the molten pool and droplet transfer in gas metal arc welding(GMAW).In this study, all the work pieces (JIS SS400) were welded at the same speed with different currents using a gas metal arc; the purpose was to observe the relationship between these factors. Using a sound capture card and microphone we collected the audio signals generated by the welding. In this paper, wavelet package transformation (WPT) decomposes the sounds signals into 128 channels within the frequency domain. The properties of the audio signals were analyzed by the wavelet-based channel energy. The ROC (Receiver Operating Characteristic) parameters, (Ac, Se, Sp, ppv, and npv), were used for classification performance evaluation. Experimental results also show that acoustic emissions are useful in the quality control of plasma arc welding.