Pattern Classification of Signals Using Fisher Kernels

Table 1

Examples of Fisher kernel applications.

Application area	Performance
Abnormal activity recognition [15]	0.857—Maximum area under curve, best performance
Handwritten word spotting [16]	89% Average precision
Shape modeling and categorization [17]	Classification error of 0.45%
Signer independent sign language recognition [18]	Average classification rate of 93.08%
Image classification [19]	Average accuracy of 60%
Spam and image categorization [20]	Overall accuracy of 79%
Web audio classification [11]	Classification accuracy of 81.79%
Object recognition [12]	Average accuracy of 93%