基于矩阵奇异值分解的信号非周期性程度指标

首先基于矩划值分解给出信号非周期性程度指标的初步定义，然后进一步提出利用最大奇异值对应的信号分量的各段之能量熵对指标进行修正的方法，并对信号奇异值分解矩阵的构造方法作了重大改进。经若干仿真信号和实测信号的测试说明，提出并改进的非周期性程度指标的性能良好。

A Signal Index: Aperiodicity Degree Based on Matrix Singular Value Decomposition

Engineering signals are all aperiodic in strict sense, and there are different aperiodicity degrees for different engineering signals. It may play an important role in signal identifications in many engineering fields to quantify the aperiodicity of engineering signals. This paper firstly introduces a rough signal aperiodicity degree index based on matrix singular value decomposition (SVD); then, the revising approach for the rough index is proposed using the energy entropy of the signal segments reconstructed from the largest singular value. Moreover, the approach to construct the matrix for SVD is improved largely. The tests using a variety of simulating signals show that the index achieved in this paper is good at indicating the aperiodicity of a signal. A real application of the index to the diagnosis of a reciprocating compressor is also made to be successful.