水下目标回波信号的分形Brown运动随机特征矢量与分类

A Method of Extracting Random Feature Vectors from Echo Signals for Classification of Underwater Targets

在研究水下目标回波信号的统计性质的基础上 ,给出以分形 Brown运动小波均方展开式中随机系数的方差作为目标回波信号随机矢量特征 ,以模糊竞争网络作为分类器对目标进行分类的方法 ;并给出了该方法的计算方法和计算步骤。大量模拟计算结果和实测目标回波信号的分类结果表明 ,新方法具有较好的实时性和较高的识别率。

The effective method of extracting the feature vectors from echo signals is the key problem to classification and recognition of underwater targets. However, the characteristic vectors in echo signals of underwater targets are often considered as nonrandom vectors (see Refs.1～3). In fact, the increments of some echo signals of underwater target possess statistical self-similarity (see Ref.4). The increment of Fractal Brown Motion (FBM) also possesses statistical self-similarity. Taking full advantage of these two statistical self-similarities, we present a new method for extracting the random feature vectors from echo signals of underwater targets based on FBM. Subsections 1.1 and 1.2 present relevant mathematical model needed for our research. Subsection 1.3 gives the five steps of calculation procedure for extracting the random feature vectors from echo signals of underwater targets with the random wavelet coefficients of FBM. Two kinds of the random feature vectors from echo signals of underwater targets based on FBM are shown in Figs.1 and 4. The classification of targets based on fuzzy neural network and Table 1 gives the results of classification based on fuzzy neural network and on only 34 samples: 20 learning samples and 14 test samples. Table 1 shows that only 2 test samples are incorrectly classified. The results of the other experiments also show that the precision of the classification of underwater targets is up to about 85%. The method based on FBM method appears to be an effective method for classifying underwater targets.