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

EURASIP Journal on Audio, Speech, and Music Processing

Table 9 Classification results for German Task. ml-MRMR was used with RCR. There are 56 non-depressed and 44 depressed subjects in the database. Best results are shown in bold. Number of selected features are shown in parenthesis for each case. The improvements are not statistically significant with respect to baseline system

Method	Classes	Precision	Recall	F1-score
*Baseline MRMR(10)*	Non-depressed	0.76	0.87	0.81
	Depressed	0.80	0.65	0.72
	Average	0.78	0.76	0.77
*Ger+Tr N*_min=3	Non-depressed	0.83	0.79	0.81
*ml-MRMR(40)*	Depressed	0.74	0.80	0.77
	Average	0.79	0.80	0.79
*Ger+Eng N*_min=5	Non-depressed	0.73	0.78	0.75
*ml-MRMR(800)*	Depressed	0.70	0.63	0.66
	Average	0.72	0.71	0.71
*GEMAPS*	Non-depressed	0.74	0.70	0.72
	Depressed	0.64	0.68	0.66
	Average	0.69	0.69	0.69