Human and dog movement recognition using fuzzy inference system with automatically generated membership functions
Chukit Ruanpeng, Sansanee Auephanwiriyakul, Nipon Theera‐Umpon · 2017
In this paper, the simple movement (walking dog, crawling human, and walking human) recognition system using the Mamdani fuzzy inference system is introduced. The membership functions of each input feature are generated automatically without experts' prior knowledges. The system produces a very high recognition rate, i.e., 93.97%, on the validation set of the cross validation. However, there are some misclassifications between walking dog and crawling human classes. The misclassifications are mainly from the incomplete segmentation of the objects of interest.