Dempster Shafer distance-based multi-classifier fusion method for pig cough recognition

High precision pig cough recognition and low computational cost is of great importance for the realization of early warning of pig respiratory diseases.Numerous researchers have improved the recognition rate of pig cough sounds to a certain extent from feature selection and feature fusion perspectives.However,there is still a margin for the improvement in the accuracy and complexity of existing methods.Meanwhile,it is challenging to further enhance the precision of a single classifier.Therefore,this study proposed a multi-classifier fusion strategy based on Dempster Shafer distance(DS-distance)algorithm to increase the classification accuracy.Considering the engineering implementation,the machine learning with low computational complexity for fusion was chosen.First,three metrics of accuracy and diversity between classifiers were defined,including overall accuracy(OA),double fault(DF),and overall accuracy and double fault(OADF),for selecting the base classifiers.Subsequently,a two-step base classifier selection approach based on these metrics was proposed to make an optimized selection of features and classifiers.Finally,the proposed DS-distance algorithm was used to fuse the selected base classifiers to create a classification.The sound data collected in the pig barn verified the proposed algorithm.The experimental results revealed that the overall recognition accuracy of the proposed method could reach 98.76%,which was better than the existing methods.This study has achieved a high recognition accuracy through ensembled machine learning with low computational complexity.The proposed method provided an efficient way for the quick establishment of high precision pig cough recognition model in practice.

A novel approximation of basic probability assignment based on rank-level fusion

Belief functions theory is an important tool in the field of information fusion. However, when the cardinality of the frame of discernment becomes large, the high computational cost of evidence combination will become the bottleneck of belief functions theory in real applications. The basic probability assignment （BPA） approximations, which can reduce the complexity of the BPAs, are always used to reduce the computational cost of evidence combination. In this paper, both the cardinalities and the mass assignment values of focal elements are used as the criteria of reduction. The two criteria are jointly used by using rank-level fusion. Some experiments and related analyses are provided to illustrate and justify the proposed new BPA approximation approach.