Parallel Big Bang-Big Crunch-LSTM Approach for Developing a Marathi Speech Recognition System

Table 2

Performance analysis of the designed framework on Marathi language with different algorithms for six different SCs in terms of error measures.

Speech corpus (SC)	Algorithms	WER	WAR	SER
SC 1	LSTM [21]	0.253333	0.826667	0.233934
	BBBC-LSTM [23]	0.233333	0.846667	0.221979
	PB3C-LSTM	0.193333	0.86	0.189498
SC 2	LSTM [21]	0.206667	0.86	0.190093
	BBBC-LSTM [23]	0.22	0.853333	0.206988
	PB3C-LSTM	0.206667	0.853333	0.231292
SC 3	LSTM [21]	0.266667	0.826667	0.26818
	BBBC-LSTM [23]	0.213333	0.866667	0.23345
	PB3C-LSTM	0.24	0.84	0.254593
SC 4	LSTM [21]	0.226667	0.853333	0.225522
	BBBC-LSTM [23]	0.226667	0.853333	0.230098
	PB3C-LSTM	0.22	0.853333	0.200145
SC 5	LSTM [21]	0.22	0.833333	0.223583
	BBBC-LSTM [23]	0.233333	0.826667	0.23908
	PB3C-LSTM	0.226667	0.866667	0.230612
SC 6	LSTM [21]	0.22	0.866667	0.215973
	BBBC-LSTM [23]	0.2	0.866667	0.216662
	PB3C-LSTM	0.18	0.88	0.194015