Skip to main content

Table 6 Regression performance of the ml-MRMR method for the English task when the minimum occurrence threshold (Nmin) is set 5. In the underlined bold case, improvement is significant compared to the baseline system. Best result with the Tr+Eng also significantly outperformed the Gemaps feature set that had an RMSE of 6.72

From: Depression-level assessment from multi-lingual conversational speech data using acoustic and text features

Dim Baseline Nmin=5 Nmin=5
   (Ger+Eng) (Tr+Eng)
3 6.85 6.66 6.66
4 6.87 7.21 7.20
5 7.31 7.78 7.78
10 7.85 7.05 7.05
15 8.08 7.94 7.70
20 7.89 7.63 7.46
40 7.25 7.12 6.63
80 7.67 6.26 6.37
100 7.57 6.38 6.15
200 7.13 6.42 6.70
400 6.95 6.72 6.73
800 6.92 6.74 6.66