Application of Transfer Learning Algorithm and Real Time Speech Detection in Music Education Platform

Table 3

Speech recognition accuracy after applying the fine-tuning of frozen layer.

Research objects	Class hours	Percentage of frozen fine adjustment layer
		10	30	50	70	90
Male/Female (double)	25	78.76	80.42	78.21	71.89	41.40
	50	78.28	78.77	78.75	72.81	50.58
	100	79.69	80.40	79.37	70.88	48.60
	200	80.20	78.81	78.86	78.33	50.70
Female/Male	25	64.50	61.22	61.52	54.52	23.72
	50	63.31	62.90	59.31	53.19	29.19
	100	63.68	63.60	61.43	54.43	34.43
	200	62.21	61.50	57.23	53.80	35.56