Cue integration with categories: Weighting acoustic cues in speech using unsupervised learning and distributional statistics

doi:10.1111/j.1551-6709.2009.01077.x

PMC full text:

Cogn Sci. Author manuscript; available in PMC 2011 Apr 1.

Published in final edited form as:

Cogn Sci. 2010 Apr; 34(3): 434–464.

doi: 10.1111/j.1551-6709.2009.01077.x

Copyright/LicenseRequest permission to reuse: Copyright notice

Table 2

Mean parameter values at the end of training for Simulations 1, 2 and 4.^b

		VOT		Voiced VL	3rd cue	VOT		Voiceless VL	3rd cue
Simulation 1	μ	1.37		188	-	51.5		178	-
	σ	5.58		44.8	-	10.2		44.8	-
	ϕ		0.493		-		0.491		-

	ρ		−0.067		-		−0.040		-

Simulation 2	μ	0.253		203	-	50.1		154	-

	σ	5.41		38.4	-	10.0		38.4	-

	ϕ	0.499		0.502	-	0.501		0.498	-

Simulation 4	μ	−0.067		204	260	50.0		155	300

	σ	5.35		38.7	10.1	10.1		38.3	10.2

	ϕ	0.501		0.498	0.502	0.499		0.502	0.498

^bValues for the 3rd cue are only applicable to Simulation 4. There is only a single ϕ and ρ for each voicing category in Simulation 1, since there is only a single value for each of these parameters in the two-dimensional Gaussians.