本文介绍一种在数字化功率谱上识别谱峰和平滑连续谱的算法。它适用于在噪声或振动谱上提取线谱特征和压缩数据量,是利用谱特征进行模式识别的有效预处理手段。这个算法的核心是在充分考虑谱的形状特点的基础上发展起来的谱峰识别逻辑。在一些实际应用中,这种方法已经取得了良好的效果。文章中给出了有关的应用实例。 This article describes an algorithm for identifying spectral peaks and smoothing continuous spectra on a digitized power spectrum. It is suitable for extracting line spectrum features and compressed data volumes in noise or vibration spectrum, and is an effective preprocessing method for pattern recognition based on spectral features. The core of this algorithm is the peak identification logic developed on the basis of full consideration of the shape and shape of the spectrum. In some practical applications, this method has achieved good results. The article gives the application examples.