Hybrid Japanese Language Teaching Aid System with Multi-Source Information Fusion Mapping

Table 4

Results of Japanese voice sequence recognition by different methods.

	ASR		Self-made datasets
	Acc	F1	Acc	F1
DNN	76%	0.84	81%	0.82
DNN-HMM	81%	0.86	83%	0.86
TDNN	89%	0.89	91%	0.89
Ours	93%	0.91	95%	0.91