Skip to main content
. Author manuscript; available in PMC: 2016 Oct 1.
Published in final edited form as: Comput Speech Lang. 2015 Jan;29(1):132–144. doi: 10.1016/j.csl.2014.02.001

Table 4.

Classification results of feature selection and parameter tuning on the TORGO database. We report both weighted and unweighted average recalls in % because of unequal proportions of the two classes in this dataset. ‘All’ is feature-level fusion with all subsystems’ features. All accuracies in %

LDA SVM

Unweighted Weighted Unweighted Weighted

Prosody 82.1 80.3 53.6 51.3
Pronunciation 94.1 93.4 83.8 82.7
Voice quality 71.9 69.0 68.4 73.4

All 93.4 92.9 53.6 51.3