Abstract: Time-frequency distribution of Choi-Williams, which is the algorithm modified based on Wigner-Vill distribution, was used to study arc voltage signals acquired during CO₂ arc welding.The signals was analyzed in time-frequency domain to get the time-frequency distribution pattern.The influence of four parameters such as analysis window type and analysis window width of time and frequency on analysis efficiency of distribution pattern was discussed on the premise of smoothing factor in kernel function of Choi-Williams algorithm.And then, the concept of information entropy was introduced.The parameters of Choi-Williams distribution were optimized by information entropy calculating which is meaningful to improve the efficiency of time-frequency analysis.The result of experiment analysis shows that the time-frequency distribution of electrical signals during welding can be expressed by Choi-Williams distribution and the principle of minimum information entropy.And then, the cross term interference is inhibited effectively and better time-frequency centralizing is obtained.