Skip to main content

Advertisement

Table 3 Classification efficiency using k-NN and SVM [%] with various combinations of dynamic parameters for acted speech (AS) and natural speech (NS)

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

Coefficients AS k-NN AS SVM NS k-NN NS SVM
MFCC 56.5 40.5 71.7 58.31
MFCC + ΔMFCC 52.7 40.5 68.7 42.91
MFCC + Δ ΔMFCC 51.1 37.13 66.9 42.91
MFCC + ΔMFCC + Δ ΔMFCC 36.7 33.75 66.9 37.5
HFCC 52.7 39.66 70.6 55.17
HFCC + ΔHFCC 47.7 37.13 67.8 40.32
HFCC + Δ ΔHFCC 46 36.7 66.9 36.23
HFCC + ΔHFCC + Δ ΔHFCC 43.9 36.7 64.6 34.59
BFCC 56.5 40.92 74.5 59.4
BFCC + ΔBFCC 56.5 36.28 70.25 44
BFCC + Δ ΔBFCC 53.6 35.86 73.4 38.41
BFCC + ΔBFCC + Δ ΔBFCC 53.6 35.86 69.9 36.28
PLP 54 21.51 71.4 36.83
PLP+ ΔPLP 52.7 16.87 66.9 22.47
PLP + Δ ΔPLP 53.6 16.87 66.9 22.47
PLP + ΔPLP + Δ ΔPLP 50.2 16.87 64.7 22.47
RPLP 55.7 43.88 70 54.76
RPLP + ΔRPLP 54.4 40.08 69.3 51.9
RPLP + Δ ΔRPLP 52.3 40.92 68.4 51.9
RPLP + ΔRPLP + Δ ΔRPLP 52.3 40.5 68 51.9
RASTA PLP 43 36.28 52.2 50
RASTA PLP + ΔRASTA PLP 40.1 35.02 51.8 36.83
RASTA PLP + Δ ΔRASTA PLP 40.9 32.48 48.5 36.83
RASTA PLP + ΔRASTA PLP + Δ ΔRASTA PLP 40.1 32.48 48.1 36.83