Revisiting Models of Concurrent Vowel Identification: The Critical Case of No Pitch Differences

Abstract

When presented with two vowels simultaneously, humans are often able to identify the constituent vowels. Computational models exist that simulate this ability, however they predict listener confusions poorly, particularly in the case where the two vowels have the same fundamental frequency. Presented here is a model that is uniquely able to predict the combined representation of concurrent vowels. The given model is able to predict listener's systematic perceptual decisions to a high degree of accuracy.

Keywords: 43.66.-x; 43.66Ba; 43.71.-k; 43.71.Es; 43.72.-p; 43.72.Qr; PACS no. 4371.An.

Figure 1.

A diagram describing our model of CVI.

Figure 2.

A scatter plot comparing the probabilities with which humans predicted concurrent vowel pairs had been presented, against probabilities predicted from the combined-class (ο; σ² = 1.03) and individual-class (×; σ² = 1.20) variants of our spectral model. The probabilities of confusing /ɝ,ɝ/ for /u.ɝ/, and correctly identifying /a,æ/, are indicated.

Figure 3.

Correlation coefficients (r) between predicted confusions for model variants, and listener confusions. ‘Sp’: Spectral pathway, ‘Te’: Temporal pathway. [11],[12],[13] are references to different cochlea filter-shapes. a) Individual classes, b) Combined classes, c) Combined classes with non-linear cochlear model [13].

Figure 4.

(Colour online) Average number of concurrent vowels correctly predicted as a function of internal noise, for variants of our model. [11],[ 12],[ 13] reference different cochlea filter-shapes.