ISCA Archive MIST 1999
ISCA Archive MIST 1999

Auditory features underlying cross-language human capabilitiesin stop consonant discrimination

Eduardo Sá Marta, Luis Vieira de Sá

For some phonemic distinctions human listeners exhibit a marked cross-language capability, in that they are capable of highly correct classification in relation to sounds (like CVs or VCVs) uttered by speakers of another language. This is particularly true regarding distinctions that are perceived in a more categorical fashion, like that of 3-way PLACE discrimination in stop consonants. It is plausible that the reason for this is a mostly common (across languages) auditory basis for human communication of this discrimination. Also, human communication of this discrimination is notably impervious to non-drastic variations in the frequencytransfer curve, which suggests that the relevant auditory features must have some inherent insensibility to these variations.

Models for two specialized auditory cells (onset cells with wide receptive fields, which can detect weak onsets synchronized across frequency, and sequence cells which detect frequency-ascending sequences composed of two onsets) were refined for the discrimination of DENTAL vs LABIAL stop consonants and applied to large spelling databases in Portuguese, German, and U.S. English. Similar discriminatory capability was observed both for German and U.S. English. Integration with a 3rd auditory feature resulted in error scores of approximately 2% when exactly the same model is applied to either German or U.S. English sounds.

Cite as: Sá Marta, E., Vieira de Sá, L. (1999) Auditory features underlying cross-language human capabilitiesin stop consonant discrimination. Proc. Multi-Lingual Interoperability in Speech Technology, 25-30

  author={Eduardo {Sá Marta} and Luis {Vieira de Sá}},
  title={{Auditory features underlying cross-language human capabilitiesin stop consonant discrimination}},
  booktitle={Proc. Multi-Lingual Interoperability in Speech Technology},