A Direct Selection Method of Feature Frequency

It is difficult to select feature peaks and frequencies to compose a feature vector because of the complication of the frequency component of the target’s acoustic signal. To deal with this problem, in this paper, all frequencies and amplitudes of spectrum curve were treated as initial data of optimization directly. The objective function was described according to the Euclidean geometrical distance of feature frequency points of two kinds of targets, and the genetic algorithm was applied during optimization. The selection result proves to be suitable for the purpose of target classification based on computer simulation examination. Thus, we developed a new method of feature frequency selection of the acoustic signal with complicated frequency spectrum.