Cross-Corpus Speech Emotion Recognition Based on Transfer Learning and Multi-Loss Dynamic Adjustment

Table 3

Experimental results of the use of ablation experiments.

Algorithm	EB	BE	EC	CE	BC	CB	Average
TLMLDA_w	51.95	31.15	31.10	30.40	32.70	53.53	38.51
TLMLDA_α	46.62	34.33	31.60	30.67	32.70	53.13	38.18
TLMLDA_L	36.76	21.12	28.70	28.01	20.05	42.71	29.56
TLMLDA_M	54.08	38.28	34.90	29.23	32.70	54.33	40.58
TLMLDA	58.93	43.16	35.40	32.74	41.00	57.11	44.72

The bold values are the highest recognition rate in each task to reflect the rationality of the TLMLDA model, because TLMLDA has obtained the best performance compared with other ablation experimental models.