信源数量估计的可视化线性聚类方法

A visual linear clustering method for estimating the number of sources

在通过传感器采集信源获得观测数据的过程中,估计信源的数量对源信号处理和观测数据分析起着非常重要的作用。为了确定稀疏信源的数量,本文提出了增强信号线性聚类特性的可视化估计方法。首先,利用短时傅里叶变换(STFT)把时域的观测信号变换成频域中的复频谱以增强观测数据的稀疏性;然后,建立一种角度余弦的相似性测度,以频谱实部分量与虚部分量之间的角度阈值来判别数据点所归属的信源;最后,把该角度阈值应用于单源点(SSP)检测中,剔除造成干扰的多源点(MSP)数据,凸显稀疏信源的线性聚类特性。实验结果表明,本文方法可以有效地增强数据的线性聚类特性,实现对信源数量直观地估计。

In the processing of acquiring the observed data by using sensors to collect sources,it is very important to estimate the number of sources for signal processing and observed data analysis.In order to determine the number of sparse sources,this paper proposes a visual estimation method to enhance the linear clustering characteristics of signals.Firstly,the short time Fourier transform(STFT)is used to transform the observed signal in the time domain into a complex spectrum in the frequency domain to enhance the sparsity of the observed data.Then,a similarity measure of angle cosine is established,and the angle threshold between the real part and imaginary part of the spectrum is used to determine the source of the data points.Finally,the angle threshold is applied to single-sourcepoint(SSP)detection to eliminate the multiple-source-point(MSP)that causes interference and highlights the linear clustering characteristics of sparse sources.The experimental results show that the proposed method can effectively enhance the linear clustering characteristics of the observed data and realize the intuitive estimate the number of sources.