Table 6 CE of a K-means classifier over the original and synthesized features
From: An evolutionary feature synthesis approach for content-based audio retrieval
FS | Basic database (6 classes) | Extended database (12 classes) | ||
|---|---|---|---|---|
Original CE (%) | Synthesized CE (%) | Original CE (%) | Synthesized CE (%) | |
Segment | ||||
STAT | 33.3 | 15.2 ± 1.6 | 54.8 | 37.4 ± 2.4 |
MFCC | 18.2 | 13.5 ± 1.7 | 36.0 | 34.4 ± 2.9 |
Δ-MFCC | 30.7 | 24.2 ± 1.4 | 48.3 | 39.9 ± 1.5 |
ΔΔ-MFCC | 37.8 | 29.1 ± 2.1 | 56.7 | 39.4 ± 2.3 |
LPC | 47.5 | 41.8 ± 1.3 | 70.8 | 68.4 ± 2.0 |
LPCC | 57.6 | 45.6 ± 1.4 | 78.9 | 74.5 ± 1.6 |
S_AUDIO | 22.1 | 18.0 ± 0.9 | 42.8 | 34.3 ± 2.0 |
Key-frame | ||||
MFCC + deltas | 37.2 | 29.2 ± 2.2 | 50.3 | 50.0 ± 2.0 |
LPC + LPCC | 75.6 | 64.4 ± 1.2 | 86.8 | 84.5 ± 1.2 |
K_AUDIO | 53.8 | 36.0 ± 4.6 | 69.1 | 46.2 ± 2.7 |