A man-machine collaborative classification mechanism based on humanware

Nowadays the machine-based classification model has made great progress, but there is still a large gap in dealing with the complex problems when compared with human beings. In addition, most existing classification models pursue high accuracy without consideration of the cost in the decision-making process. To address these issues, a man-machine collaborative recognition mechanism based on humanware system is proposed. By dividing the workspace of experts and machines reasonably, machines handle the samples that are easy to be distinguished, and leave the confusing samples to the experts. According to the proposed man-machine collaborative mechanism, the support vector machine (SVM) is modified. Finally, the effectiveness of the improved SVM model is verified by the experiments on Caltech 256 dataset.