论文部分内容阅读
本系列文章的工作是在舰船噪声谱图的基础上，利用模糊神经网络对舰船进行分类识别．本文是系列文章的第二篇，研究如何用线谱模板来记忆特定舰船的稳定线谱特征和涉及的一些问题．首先调查线谱分析参数──平均时间、平均次数对线谱稳定出现的影响，比较了两个不同平均次数对线谱稳定出现的影响，稳定线谱（出现率＞70%）和不稳定线谱的比例，说明了利用稳定线谱作识别时需作长时间的平均．研究了利用稳定线谱建立舰船线谱特征模库时所用的统计方法，稳定性的限定和有关参数的定义．在43条舰船65种情况1000多个样本（原始记录时间总长约为3．5小时）中调查了线谱稳定性和稳定线谱的唯一性，调查结果说明唯一性在统计意义下成立，其平均重叠率为5％，无稳定线谱的舰船占8％．对于线谱数量丰富程度的特征，不存在甲类舰船线谱丰富，而乙类舰船线谱不丰富的规律．
The work of this series of articles is based on the ship noise spectrum, the use of fuzzy neural network classification of the ship classification. This article is the second in a series of articles that explores how line spectrum templates are used to memorize the stability characteristics of a particular ship and some of the issues involved. First of all, the influence of line analysis parameters, such as average time and average number of times, on the stability of line spectra was investigated. The effects of two different average times on the stability of line spectra were compared. The stable line spectrum (occurrence rate> 70%) and unstable line The ratio of spectra shows the average length of time required to make use of the stable line spectrum for identification. The statistic method, the definition of stability and the definition of the relevant parameters used in the establishment of ship line spectral feature database using the stable line spectrum are studied. The uniqueness of the line spectrum stability and the stable line spectrum was investigated in over 1000 samples of 65 cases of 43 ships (the original recording time was about 3.5 hours in total). The survey results show that the uniqueness is established statistically, The average overlap rate of 5%, non-stationary line spectrum of ships accounted for 8%. For the characteristics of abundance of line spectrum, there is no regularity that the class A ships are rich in spectrum but the class B ships are not rich in spectrum.