摘要
在水声信号处理中，ＤＥＭＯＮ和ＬＯＦＡＲ已被证明为有效的方法，特别是对微弱信号的检测和目标的识别和分类。有的时候，我们还需要知道接收信号频谱的细微结构．一般说来，只有长的时间数据才有可能得到高的频率分辨力，但是由于实际系统软、硬件方面的限制，这样作并不总是可能的．如果我们只是对某些频率附近的谱结构感兴趣，那么ＺｏｏｍＦＦＴ就是一种解决高分辨率谱分析的折中方法．已有的讨论ＺｏｏｍＦＦＴ的文献大体可以分为两大类，即复包络解调ＺｏｏｍＦＦT（Ｃｏｍｐｌｅｘｍｏｄｕｌａｔｉｏｎ）［４］和级连ＦＦＴ（ｃａｓｃａｄｅＦＦＴ）［５，６］．前者需要对输入信号进行复解调、低通滤波、降采样等一系列繁复的操作．而后者通常利用前后两次ＦＦＴ，经过相位和幅度修正得到所需频段的细化谱估计，因而易于实现，可作为一种有效的窄带处理器。本文在给出级连ＦＦＴ法ＺｏｏｍＦＦＴ理论推导的基础上，试图探讨其与复包络解调法之间的内在对应关系，并分析了窗函数、采样率、重叠率等参数的选取对估计结果的影响，最后给出一种简化的ＺｏｏｍＦＦＴ算法，它可以大大缩短实时数据的运算次数．并给出了系统模拟的结果。
DEMON and LOFAR have been proved the powerful mean in underwater acoustic signal processing, espe-cially in weak signal detection and target noise classification. Sometimes one need to know the fine structure of frequencyspectrum of received signal. It is necessary to take a very long data to get high frequency resolution. This is not alwayspossible due to the hardware and software limitation. ZoomFFT is one of the trade-off consideration for solving highfrequency resolution problem, if we are only focus on some special frequency bins. Previous discussions mainly bifurcate into two different representations, the Complex Modulation[4] and Cascade FFT1516]. The former one traditionallyneeds some kind of special treatment, e.g. the complex modulation, Lowpass filtering, down--sampling. While the latterachieves the same result by two successive FFT, with necessary modifications in phase and amplitude, thus is feasiblefor real-time implementation. Based on some theoretical analysis, a relationship between the Complex Modulation andCascade FFT has been described in this paper. In addition, the selection of parameters such as sample rate, overlapfactor has been discussed. Finally, the algorithm is presented and some simulation results are illustrated.
出处
《声学学报》
EI
CSCD
北大核心
2000年第2期129-133,共5页
Acta Acustica
