Skip to main content

Advertisement

Table 4 Accuracy performance of emotion recognition [%] without selection (–) in comparison with two different selection methods (FMS and SFS) and extraction (S-PCA) methods using k-NN algorithm as a classifier

From: Efficiency of chosen speech descriptors in relation to emotion recognition

  Acted speech Natural speech
  SFS FMS S-PCA SFS FMS S-PCA
f 0 31.6 38.4 40.92 32.9 39.8 40.9 46.58 40.9
f 1f 3 38.8 39.2 38.8 34.2 39.5 39.6 45.73 37.1
Energy 30.4 36.7 36.7 31.6 47 47.7 55.96 44.6
LPC 36.7 36.7 36.7 36.3 57.8 57.8 63.06 48.5
BFCC 55.7 64.4 64.4 51.5 76.7 77.7 76.7 71.1
MFCC 56.5 59.9 58.10 40.5 71.7 74 74 62.7
HFCC 51.1 52.7 52.7 45.6 70.4 72.9 75.61 68.8
PLP 54 54.4 57.25 51.9 71.4 71.4 71.4 70.4
RPLP 55.7 58.6 59.9 51.9 70 71.3 71.38 64.2
RASTA 43 46.8 49.36 35 52.2 52.2 72.76 44.1