Forensic Speaker Comparison Using Evidence Interval in Full Bayesian Significance Test

Table 1

Occurrence of the scenarios described in Figure 4 for interval inference methods.

		Percent of occurrences (%)
Scenario	SNR (dB)	Gosset	Morrison	FBST	Punctual
(a) Correct classification	12	84.4	84.3	84.0	85.6
	15	88.6	88.6	88.4	89.6
	17	91.0	91.0	90.7	91.7
	20	93.7	93.7	93.5	94.2
	23	95.4	95.5	95.3	95.9
	25	96.4	96.4	96.3	96.8
(b) Middle interval	12	2.4	2.6	3.2
	15	1.8	1.9	2.3
	17	1.5	1.5	1.9
	20	1.0	1.0	1.4
	23	0.9	0.9	1.2
	25	0.7	0.7	0.9
(c) Miss classification	12	13.2	13.1	12.8	14.4
	15	9.6	9.5	9.3	10.4
	17	7.5	7.5	7.4	8.3
	20	5.3	5.3	5.1	5.8
	23	3.7	3.6	3.5	4.1
	25	2.9	2.9	2.8	3.2
Average interval length (Np)	0.074	0.080	0.092