非高斯背景下的频域广义匹配检测

Generalized matched detection of non-Gaussian background in the frequency domain

采用广义匹配滤波器在功率谱未知的非高斯噪声中检测宽带信号。在估计功率谱时,为了减小周期图法谱估计产生的误差,采用平滑周期图法来估计非高斯噪声的功率谱,并在此基础上对信号进行恒虚警概率损失下的广义匹配检测。实验表明,用平滑周期图法估计功率谱并进行的广义匹配检测,其检测性能稍弱于功率谱取真值时的检测,但远远优于传统的匹配滤波器。

Detection of a known broadband signal in non-Gaussian noise of unknown power spectral uses generalized matching techniques. In order to reduce variance of power spectrum estimate using the periodogram, we use the smoothed periodogram to obtain the estimated spectral of the non-Gaussian to replace the true spectral, and process the signal to do generalized matched filtering under the loss of a constant false alarm rate. Experiments show that performance of the generalized matched detection that estimates power spectrum with smoothed periodogram is only slightly inferior to the true power spectrum, and much better than the result obtained only using a match filter.