一种恒虚警线谱检测算法研究

提出基于高低分辨率功率谱匹配度估计的线谱检测算法,该算法对信号分别进行高分辨率功率谱估计和多段低分辨率功率谱估计,然后以高分辨率功率谱估计为模板,用多段低分辨率功率谱估计进行校验的方式进行线谱检测。基于单频信号加高斯白噪声假设条件,理论上导出该检测方法的检测率和虚警率,并进行了仿真验证。与传统的线谱检测方法相比,该线谱检测算法不需要预知噪声功率谱,且具有恒虚警的特点。在车辆目标远距离声学检测应用中,该检测算法具有较好的检测性能。

An algorithm of line-spectrum detection with constant false alarm rate

A new line-spectrum detection algorithm is proposed based on the matching extent of high-resolution and low-resolution power spectrums. For a certain data section, the high-resolution power spectrum is firstly estimated as a template, and then several low-resolution power spectrums are estimated and compared with the template to verify the line-spectra. Under the condition of sinusoid signal mixed with Gaussian white noise, the detection rate and the false alarm rate of this algorithm is deduced theoretically and proved by the Monte-Carlo simulation. In comparison with the conventional methods, this algorithm doesn＇t need any information of the noise power spectrum and has constant false alarm rate. In the application of vehicle long-distance detection, this method achieves satisfying performance.