一种基于Otsu准则的盲信噪比估计算法

A blind SNR estimation algorithm based on Otsu criterion

针对基于子空间分解信噪比估计算法中信号子空间维数估计不准确的问题，提出了一种基于Otsu准则的盲信噪比估计算法。该算法构造接收信号的自相关矩阵，利用图像分割中的Otsu准则将含噪信号空间精确分成信号子空间和噪声子空间，求得信噪比。仿真结果表明，信噪比为3～30dB时，高斯信道下常用调制信号的信噪比估计标准差小于0．05dB，与基于最小描述长度（minimumdescriptionlength，MDL）准则、Akaike信息准则（Akaikein—formationcriterion，AIC）等算法相比能适应更短的数据长度。

In order to solve problems of low precision of estimating dimension of signal subspace in the algorithm of SNR es- timation based on subspace decomposition, an efficient algorithm is proposed to estimate SNR. Firstly, an autocorrelation matrix of received signal must be constructed, then we can divide the signal space that contains noise into signal subspace and noise subspace by the use of Otsu Criterion that is widely applied in image segmentation. Finally we can obtain SNR es- timation. Computer simulations show that the standard deviations for different modulation signals are less than 0.05 dB when the actual SNR ranges from 3 dB to 30 dB. And this algorithm performs better in shorter data samples environment compared with minimum description length （MDL） and Akaike information criterion （AIC） methods.