Feature extraction of underwater low-velocity targets based on 1 1 over 2-spectrum

The feature extraction and target recognition of low-velocity targets near the port is difficult in signal processing. After the spectral characteristics of low-velocity targets are analyzed, the paper uses 1(1/2)-spectrum as the method of feature extraction. The advantages are as follow. 1(1/2)-spectrum can suppress the Gaussian white noise. It can enhance weak fundamental frequency components of harmonic signal and extract quadratic phase coupling components. In view of the advantages mentioned above, the paper applies 1(1/2)-spectrum based on higher order statistics to the feature extraction of low-velocity targets for the first time. A 20-dimensional feature vector is extracted by using 1(1/2)-spectrum sub-band energy method. High and low velocity targets are classified with the use of BP neutral network. The results show that the average recognition rate is more than 85%. It is concluded that the method of 1(1/2)-spectrum is effective to distinguish high-velocity targets with low-velocity ones.