光度信号的频域模式识别方法研究

Modal identification of luminosity signals in frequency domain

通常目标的光度信号信噪比小 ,测量误差大 ,难以目标识别。本文采用数字信号处理和模式识别相结合的方法对目标光度信号进行识别。先将经过预处理的光度信号通过一种类似于Gabor变换的改进的离散傅里叶变换 (DFT)转换到频域 ,再滤波、划分频带 ;然后将选出的小频带内能量作为特征向量 ,在频域引入Fisher判别 ,用已知类别数据得出判别函数矩阵 ,建立判别规则 ;最后将未知类别的数据代入判别函数矩阵 ,通过判别规则完成识别。成功地对三颗卫星的光度信号进行了判别 ,克服了其光度信号信噪比小、测量误差大。

Object luminosity signals were identified by means of digital signal processing(DSP) and modal identification. Modified luminosity signal was transferred to frequency domain by special discrete fourier transfer(DFT) which is like gabor transfer. Then, signal was filtered and partitioned into fine frequency bands. After that, Fisher's discriminance was used in frequency domain wth characters, which were energies of a group of selected fine frequency bands. The identification function matrix(IFM) was obtaind through recognized data and identification rulers were given. After unidentified data had been put into the IFM, they were categorized under the rulers. Three satellites' luminosity signals were identified successfully and difficulties like low SNR, abnormal errors from measurement, complex model of noise were overcome.