传统的地震数据采集方式是炮与炮之间分开激发，施工效率较低，多震源地震混采技术通过多炮同时或延时激发，大幅提高了施工效率，但同时也给地震数据带来了较大的混合噪声，传统方法的去混效率偏低，为此，将曲波变换与快速阈值迭代法进行结合，在此基础上使用指数阈值模型，得到一种基于稀疏反演快速高精度的多震源数据分离方法。在曲波域分离过程中，采用软阈值函数和指数阈值模型，经过迭代获得了较好的分离结果，此外，还对其抗噪性能进行了相关的研究。理论数值模拟和实际数据的应用，表明本文方法相较于传统的分离方法，收敛速度更快，保留了更多的有效信号，具有更高精度的分离效果，且有良好的抗噪性能。

The traditional method of seismic data acquisition involves shots being separately excited, which results in low construction efficiency. Construction efficiency is significantly improved by the multi-source seismic mixed acquisition technology through the simultaneous or delayed excitation of multiple shots. However, substantial mixed noise is introduced to the seismic data by this technology, and the demixing efficiency of traditional methods is relatively low. To address this issue, the combination of the curvelet-transform with the fast threshold iteration method was used, and an exponential threshold model was applied to develop a fast and high-precision method for separating multi-source seismic data based on sparse inversion. During the curvelet domain separation process, a soft threshold function and exponential threshold model were utilized, resulting in favorable separation results after iterations. Additionally, the method's noise resistance performance was investigated. Theoretical numerical simulations and the application of field data demonstrated that the proposed method exhibits faster convergence, retains more effective signals, achieves higher-precision separation, and shows good performance on noise resistance compared to traditional separation methods.