中文摘要：

奇异值是矩阵的一个主要特征，通过对矩阵奇异值分解的理论分析，描绘了利用奇异值分离周期性信号的具体方法。针对单缸柴油机排气噪声主要由以点火频率为基频的一种周期噪声的特点，作为矩阵奇异值分解技术的一个应用，对用声传感器采集的排气噪声信号进行分解和重构，提取周期信号，并利用短时傅里叶变换，从频域提取信号特征，为排气消声器结构参数的选取提供理论依据。

英文摘要：

Singular value is one of the main characteristic of matrix. In this paper, a method to sep- arate periodic signals is elaborated through the theoretical analysis of matrix singular value decomposition. Aiming to the characteristic that the exhausting noise of single-cylinder diesel engine is mainly a periodic noise with sparking frequency being the basic frequency, the exhausting noise signals acquired by sound sensor are resolved and then reconstructed, and the periodic signals are extracted. Based on short time Fourier transform, the character of signal is extracted from the frequency domain.