An information model and method of feature fusion

For a specific pattern recognition problem, many kinds of feature information can be extracted by different signal analysis means. How to use efficiently such kinds of feature information is of wide concerned in the field of pattern recognition. This paper presents an information network model that considers the algorithms of feature extraction, feature fusion and classification as information engines. A measuring criterion of feature fusion is proposed by analyzing the feature fusion mechanism. In addition, a fusion method based on dynamic programming is presented. In the sense of dynamic programming, the complex process of obtaining the global satisfactory solution could be dramatically simplified. The application in the classification of underwater acoustic signals obtained satisfactory result