Table 3.

Performance comparison of proposed approach with SOTA approaches. CE_l refers to models with LSTM block.

	DAIC-WOZ				Vocal mind
Approach	$F_1(D)$	$F_1(H)$	BAc.	RMSE	$F_1(D)$	$F_1(H)$	BAc.	RMSE
Sequence	0.32	0.70	0.51	7.41	0.32	0.67	0.49	7.63
eGeMAPS	0.32	0.71	0.52	7.05	0.27	0.74	0.50	7.22
FVTC-MFCC	0.37	0.79	0.58	6.41	0.30	0.77	0.54	6.85
FVTC-FMT	0.39	0.79	0.59	6.37	0.34	0.76	0.55	6.82
Mk-CNN (COVAREP)	0.35	0.70	0.52	7.39	0.30	0.68	0.49	7.61
LSTM (OpenSMILE)	0.39	0.73	0.56	6.82	0.34	0.75	0.55	6.94
MK-CNN (ECAPA-TDNN)	0.43	0.78	0.60	6.35	0.32	0.80	0.56	6.64
LSTM (ECAPA-TDNN)	0.46	0.79	0.63	6.31	0.34	0.81	0.57	6.62
CE${}_l$ (ECAPA, COVAREP)	0.47	0.80	0.64	6.19	0.37	0.81	0.59	6.51
CE${}_l$ (ECAPA, OpenSMILE)	0.51	0.83	0.66	6.01	0.43	0.84	0.64	6.28