Dielectric Spectrum Feature Vector Extraction Algorithm of Ground Penetrating Radar Signal in Frequency Bands

This letter presents a new feature extraction algorithm for ground penetrating radar (GPR) data. Our new algorithm can fully utilize the dielectric spectrum of the GPR data in different frequency bands. First, according to power spectral distribution of impulse source signal, we select the frequency band of GPR data´s main energy spectrum. Second, we divide the frequency band into different subbands by bandpass filters groups. The features of “slow-decay point” and “fast-decay point” are separately extracted from each subband, and encoded to build the feature vector. The experimental results show that the classification results used in the proposed feature vector agree well with the parameter values set during experiments. The proposed algorithm can be applied on the targets classification of the GPR signals. In addition, it has advantages of low time complexity and strong antinoise ability.