欢迎访问《兵工学报》官方网站,今天是 分享到:

兵工学报 ›› 2014, Vol. 35 ›› Issue (10): 1630-1637.doi: 10.3969/j.issn.1000-1093.2014.10.017

• 论文 • 上一篇    下一篇

差分二次平均修正的频域相位补偿线谱检测方法

戴文舒1,2, 陈亚3, 陈新华1, 孙长瑜1, 余华兵1   

  1. (1.中国科学院 声学研究所, 北京 100190;2.中国科学院大学, 北京 100190;3.海军驻无锡地区军事代表室, 江苏 无锡 214061)
  • 收稿日期:2013-08-06 修回日期:2013-08-06 上线日期:2014-11-28
  • 通讯作者: 戴文舒 E-mail:dws.01@163.com
  • 作者简介:戴文舒(1987—),女,博士研究生
  • 基金资助:
    国家自然科学基金项目(61372180)

An Algorithm of Line-spectrum Detection Using Frequency Domain Phase Compensation with Difference and TPM Technique

DAI Wen-shu1,2, CHEN Ya3, CHEN Xin-hua1, SUN Chang-yu1, YU Hua-bing1   

  1. (1.Institute of Acoustics, Chinese Academy of Sciences, Beijing 100190, China;2.University of Chinese Academy of Sciences, Beijing 100190, China;3.Navy Military Representative Office in Wuxi Area,Wuxi 214061,Jiangsu, China)
  • Received:2013-08-06 Revised:2013-08-06 Online:2014-11-28
  • Contact: DAI Wen-shu E-mail:dws.01@163.com

摘要: 针对常规平均功率谱方法幅度平方后相位信息丢失造成检测性能损失问题,通过对观测数据分段进行快速傅里叶变换后的统计特性进行分析,提出了高斯噪声背景下检测未知线谱的频域相位补偿方法。采用蒙特卡洛法估计相位补偿因子,先进行差分法去除零星野值,然后使用二次平均(TPM)去除连续野值,最后使用估计的相位补偿因子均值构造广义似然比检验统计量实现检测。从理论上对比了广义平均周期图(AVGPR)法、广义功率谱(GPR)法及文中方法的检测性能。仿真结果表明,文中方法充分利用信号的相位信息,参数估计简单,相位补偿因子估计准确,在一定虚警概率下较AVGPR法的检测性能提高了接近5 dB左右。

关键词: 声学, 快速傅里叶变换分析, 线谱检测, 频域相位补偿, 检验统计量, 功率谱

Abstract: The common average power spectrum estimation algorithm may cause the loss in detection performance because power spectrum amplitude is simply squared. The statistical properties of FFT of the segmented observations are analyzed, and a novel algorithm using frequency domain phase compensation for detecting line-spectrum in Gaussian noise is proposed. The phase compensation factor is estimated using Monte Carlo method. The sporadic and continuous outliers are removed by using difference and TPM method. Finally the statistics is checked by using generalized likelihood ratio. The detection performances of AVGPR, GPR and the proposed algorithm are theoretically compared. The simulation results show that the detection performance of the proposed algorithm is improved by 5 dB under a certain false alarm rate compared to AVGPR algorithm. It fully utilizes the phase information of the signal and has the advantages of simple and exact parameter estimation and good robustness.

Key words: acoustics, FFT analysis, line-spectrum detection, frequency domain phase compensation, test statistics, power spectrum

中图分类号: