Research Of Target Identification Algorithm Based On UWB Signal And Diagonal Slice Of The Complex Bispectrum

In recent years, ultra wideband (UWB) technology is one of the hot research spots in the field of short-range wireless communication technology. UWB signal does not need a carrier. The structure of the system is simple. Using time-hopping spread spectrum, the radio frequency bandwidth of UWB signal can reach more than1GHz. UWB signal has an anti-multipath capability along with good concealment. UWB signal also has a strong ability to penetrate obstacles, such as leaves, water, land, concrete, etc. It can realize the target identification under foliage environment.Take advantage of the UWB signal technology mentioned above, we put forward a new feature extraction method base on diagonal slice of the complex bispectrum to solve the target identification problem under foliage environment, It achieves communication process and target identification at the same time.The basic principle of target identification is to identify the target using the feature information of UWB radar echo, such as amplitude, phase, polarization, etc. Then output the final decisions according to the discriminant function determined by a large number of training samples. There are several steps including information collection, data preprocessing, feature extraction and classification in a relatively complete pattern recognition system. The proposed algorithm mainly study and make improvement in two aspects of feature extraction and classification. In practice, the UWB radar echo signal is a typical non-Gaussian process with colored noise, which is especially suitable for higher-order statistics to solve. The diagonal slice of bispectrum can obtain the amplitude and phase information of the nonlinear signal. Then Chirp-z transform is used to computed the feature vectors. It can choose the selected arbitrary transform paths and resolution by adjusting the parameters. The Chirp-z transform enlarge the diagonal slice of bispectrum in the frequency domain into the z-plane, so the separability feature beyond the bispectrum can be obtained, which can increase the flexibility of the analysis. Since each feature vector contains different information, to compress feature dimensionality and select optimal separability features, the extended Fisher measure is presented.Then apply the ant colony optimization algorithm based on grid strategy to optimize the parameters of support vector machine (SVM). Finally SVM is used to classify the targets. According to experiment data, the results show that it is quite effective for target identification. It has a certain significance and practice value in target identification under the foliage environment.