. Author manuscript; available in PMC: 2021 Mar 18.

Published in final edited form as: IEEE/ACM Trans Audio Speech Lang Process. 2017 Nov 27;25(12):2281–2291. doi: 10.1109/taslp.2017.2759002

TABLE I.

Confusion matrix between prompt and perception labels for breathy (B), modal (M), strained (S) and rough (R) qualities. Overall agreement was 71.2% (κ = 0.6). Utterances with “pure” voice quality are in bold.

		Perception
		B	M	S	R
Prompt	B	164	8	0	0
	M	21	311	11	7
	S	7	86	93	8
	R	43	5	69	88
	Pure	178	410	108	38