From: Hypothesis testing for evaluating a multimodal pattern recognition framework applied to speaker detection
Test 1
Test 2
Input features
MFCCs mean
Optimized audio features
AUC
0.88
0.92
0.75
0.84
Accuracy
84, 6%
86, 7%
73, 4%
85, 1%
η
0.14
0.18
0.10
0.19