Abstract: Because arc sound signal is easily disturbed by environmental noise in MIG welding,the characteristics of arc sound signal is influenced,and so the filtering noise is essential condition of using arc sound signal.In the experiment,the arc sound signal and voice noise signal were mixed into two random observation signals.The fast fixed-point algorithm of independent component analysis (ICA) was used to separate the noise from random observation signals.The arc sound signal and the separated signal were analyzed in time domain and frequency domain.The experimental results show that the ICA can be used to separate environmental noise from arc sound signal.The ICA separating technology provides technical basis for extracting characteristics of arc sound signal and monitoring welding quality via arc sound signal.