Detecting Depression Using an Ensemble Logistic Regression Model Based on Multiple Speech Features

Table 3

Classification outcomes of each individual classifier for males.

Features	SVM			GMM			LR
	Sen. (%)	Spe. (%)	Acc. (%)	Sen. (%)	Spe. (%)	Acc. (%)	Sen. (%)	Spe. (%)	Acc. (%)
MFCC	56.14	64.91	60.66	62.72	58.22	60.40	62.50	60.75	61.60
PROS	61.96	70.39	66.30	61.75	74.14	68.13	63.15	71.10	67.24
SPEC	63.36	73.94	68.81	65.84	71.60	68.81	67.35	70.69	69.07
GLOT	36.32	60.95	49.01	47.95	54.26	51.20	44.07	54.56	49.48
MFCC + PROS	60.67	69.78	65.36	63.69	70.49	67.19	65.41	68.36	66.93
MFCC + SPEC	59.05	72.72	66.09	64.55	69.17	66.93	63.58	69.07	66.41
MFCC + GLOT	53.56	66.53	60.24	61.96	60.65	61.29	61.10	60.75	60.92
PROS + SPEC	63.25	73.83	68.70	62.72	74.14	68.60	67.13	72.21	69.85
PROS + GLOT	60.99	71.60	66.46	61.96	72.92	67.61	62.82	71.20	67.14
SPEC + GLOT	62.61	75.15	69.07	65.19	70.89	68.13	66.70	70.99	68.91
MFCC + PROS + SPEC	60.99	72.92	67.14	65.63	72.31	69.07	64.55	70.39	67.56
MFCC + PROS + GLOT	59.59	72.62	66.30	63.36	69.27	66.41	62.82	67.24	65.10
MFCC + SPEC + GLOT	60.24	72.82	66.72	65.84	69.98	67.97	64.66	68.66	66.72
PROS + SPEC + GLOT	60.02	73.83	67.14	62.82	74.24	68.70	64.12	71.91	68.13
MFCC + PROS + SPEC + GLOT	61.85	73.12	67.66	66.27	71.30	68.86	64.33	72.21	68.39

Maximum of sensitivity (sen.), specificity (spe.), and accuracy (acc.) is shown in bold.