Audio-Textual Emotion Recognition Based on Improved Neural Networks

Table 1

The comparison of the experimental results.

Model	Accuracy (%)
	Angry	Excite	Neutral	Sad	WA	UA
S_DNN	54	39	38	64	47.37	48.75
S_CNN	49	29	46	74	49.26	49.50
S_CL	65	40	48	59	52.23	53.00
S_CBLA	44	51	60	69	56.55	56.00
T_LSTM	68	59	65	61	63.56	63.25
T_BL	71	65	69	52	63.70	64.25
M_D_BL	75	61	49	86	65.59	67.76
M_CL_BL	74	70	60	64	66.26	67.25
M_CL_BLR	72	70	67	73	70.18	70.50
M_CBLA_BL	77	75	57	77	69.37	71.50
M_CBLA_BLR	74	65	65	81	70.40	71.25