Skip to main content

Advertisement

Table 9 Feature selection for pitch classification with auditory model and DFT: number of selected features and error rates

From: A computational study of auditory models in music recognition tasks for normal-hearing and hearing-impaired listeners

Method No selection Channel groups Feature groups
   Forward Backward Forward Backward
RF: number of features 41 × 29 = 1189 4 × 29 = 116 35 × 29 = 1015 41 × 2 = 82 41 × 28 = 1148
RF: error rate 0.08 0.10 0.07 0.09 0.08
SVM: number of features 41 × 29 = 1189 5 × 29 = 145 23 × 29 = 667 41 × 2 = 82 41 × 9 = 369
SVM: error rate 0.07 0.10 0.07 0.09 0.07