大气垂直探测仪在进行光谱定标或者光谱分析时，对局部光谱的光谱分辨率有较高的要求，这时需要对光谱进行细化处理。传统的光谱细化，多采用DTFT、DCT 等方法，计算量非常大，即使采用优化的CZT 算法，其计算量也是非常巨大的。随着市面上大量多核处理器的出现，使得利用多核处理器对CZT 算法进行快速并行优化成为可能。本文介绍了一种基于多核并行实现CZT 的算法并分析了算法运算量，在TMS320C6678 多核DSP 处理器上进行了验证。实验结果表明，该方法可以极大地缩短CZT 的运行时间，同时可以通过调整参数控制对底层资源的使用，为实现大点数的频谱细化提供了一种全新的解决方案。

The spectral calibration and analysis of an vertical atmospheric sounding instrument requires high spectral resolution. Therefore, a refinement operation is needed in such cases. Traditional spectral refinement uses methods that require extensive calculations, such as discrete-time Fourier transform, discrete cosine transform, and others. Even the optimized chirp-Z transform (CZT) algorithm is computationally heavy. However, the many multi-core processors that have recently appeared on the market have made the rapid parallel optimization of the CZT algorithm possible. This study proposes a parallel CZT algorithm and analyzes its computations. The proposed algorithm is verified using the TMS320C6678 multi-core DSP platform. The experimental results show that this method can significantly shorten the running time of the CZT. In addition, developers can adjust its parameters to control its use of calculation resources. The proposed algorithm provides a new solution for realizing the spectral refinement of large points.