Skip to main content

Advertisement

Table 3 Within-corpus results for binary classification on the USC EMA database

From: Articulation constrained learning with application to speech emotion recognition

Vowel SVM ACO ACL Best target Best group BCUAR
Arousal
/AA/ 95.26 95.12 93.46 93.84 (JAW, VEL, Y) 93.51 (JAW) 54.51
/AE/ 91.09 91.47 91.88 91.74 (LIP, POS, X) 92.01 (LIP) 53.16
/IY/ 95.34 95.59 95.74 95.57 (JAW, POS, X) 95.80 (JAW) 53.35
/UW/ 94.49 93.64 93.85 93.64 (TNG, ACC, X) 94.41 (LIP) 57.75
FULL 99.51 99.51 99.13 99.64 (LIP, POS, X) 99.25 (TNG) 52.88
Valence
/AA/ 78.38 77.89 79.46* 79.38 (JAW, ACC, Y) 79.34 (LIP) 53.17
/AE/ 76.75 78.40 78.34 78.26 (TNG, VEL, X) 78.20 (JAW) 52.47
/IY/ 70.14 69.45 73.04* 73.90 (LIP, VEL, Y) 73.01 (LIP) 54.02
/UW/ 76.25 72.01 76.13* 77.63 (JAW, ACC, X) 76.30 (LIP) 53.44
FULL 92.85 92.79 95.11* 95.33 (TNG, VEL, X) 95.31 (TNG) 52.68
  1. The UAR is expressed in percentage. The columns of the table represent results from several models: support vector machine (SVM), acoustic only model (ACO), articulation constrained learning (ACL), best target using ACL, best group of targets using ACL, the by-chance unweighted average recall. *Statistically significant improvement(p<0.05) over compared methods