Afan Oromo Speech-Based Computer Command and Control: An Evaluation with Selected Commands

Table 2

Phone-based models’ recognition.

Parameter tuning values		Percent of Word monophone			Percent of Word internal triphones			Percent of Word tied-state tri-phones recognized
-p options	-s options	Correctly recognized	Accurately recognized	Word error rate (%WER)	Correctly recognized	Accurately recognized	Word error rate (%WER)	Correctly recognized	Accurately recognized	Word error rate (%WER)
0.0	3.0	83.02	78.12	21.88	91.51	88.99	11	90.32	86.87	13.12
0.2	5.0	83.02	78.12	21.88	91.51	88.99	11	90.32	86.87	13.12
0.4	7.0	83.02	78.12	21.88	91.51	88.99	11	90.32	86.87	13.12
0.6	9.0	83.02	77.98	22	91.51	88.99	11	90.32	86.87	13.12
0.8	11.0	83.02	77.98	22	91.51	88.99	11	90.32	86.87	13.12
1.0	13.0	83.02	77.98	22	91.51	88.99	11	90.32	86.87	13.12
1.2	15.0	83.02	77.85	22.1	91.51	88.99	11	90.32	86.87	13.12