A Method Of Object Identification Based On Gabor-Network And UWB

As a short-distance wireless communications technology, Ultra-WideBand (UWB) technology has a distinct advantage in transmission efficiency, power and price cost. Therefore, it has become the focus of academia and industry. Because of its low power consumption, high resolution ability and strong multipath penetration, UWB has a huge application value in the field of personnel resolution, object recognition and obstacle detection, such as disaster relief, security testing and battlefield reconnaissance.In this paper we take the advantages of UWB’s environmental awareness ability. With the help of that UWB signal is able to complete the target recognition features while its communication process, we establish different test scenarios in the outdoor environment for the target recognition test. In these scenarios, different target objects are placed in the same position, otherwise the same target object are placed in different position respectively. We process the UWB signal sending and receiving test under those scenario.Because of the different communication environment, UWB received signal will be changed and appear the corresponding difference as the communication channel contains different objects. After getting enough sample data, we process the Gabor transform of the received UWB signal, extract the Gabor coefficients of them, and then combine it with the neural networks. At last, we use the Gabor coefficients as the input vectors of the neural networks and make the classification result as the output of neural networks.The neural networks will be used after completing self-study by training samples, with its learning and classification ability so we can achieve the target object classification and localization.Further, in order to study Gabor neural network performance of target recognition under the low SNR condition, we set up a theory model of indoor transmission environment, on the basis of which we generate the simulation signal with strength of difference Gaussian white noise.In order to achieve the target recognition under the condition of low SNR, the Gabor neural network recognition method was improved and optimized in this paper, namely a denoising processing unit is added between the Gabor transform and neural network We select a appropriate Gabor threshold denoising algorithm to filter out the noise component.Then we input the signal’s Gabor coefficient after denoising to the neural network to identify the target. The final result of simulation shows that the improved Gabor neural network’s recognition accuracy can reach more than90%, which means the Gabor neural network realized target recognition under the condition of low SNR...