Vector phase space for speech analysis via dimensional analysis

H. M. Hubey

Research output: Contribution to journalReview articlepeer-review

2 Scopus citations


A vector space using dimensional analysis is produced in which one can show all the phonemes/phones of all languages. Vowels, and consonants can all be shown in this phase space. Furthermore, the three-dimensional vector space for vowels, which in simplified form can be shown to be related to the distinctive features, can also be compressed to fit in this phase space for speech. This phase space can be shown to be both based on articulatory/geometric considerations, that is the two-tube model of Fant and Stevens, and also on the quality/perception arguments based on formant studies (Peterson & Barney, and Clark & Yallop). It can be used to clarify and unify many linguistic phenomena such as child language (Anderson, Jacobson), aphasia, sonority, the cardinal vowel diagram (Jones, Ladefoged), diphthong trajectories (Carre & Mrayati). It is shown that the sonority scale is directly correlated with this space in that sonority is related to the distance of the phones/ phonemes from the origin. Hence sonority is a function of the magnitudes of the vectors (phonemes/phones) of this space. Diphthong and vowel confusion that crops up when using Artificial Neural Networks (Kohonen) for vowel recognition is easily explicable in this space. The fortition-lenition phenomena and phonological strengths (Foley) are nothing but vector phenomena in this space. The reasons that almost all languages have the phonemes /ptskn/ can be clearly shown in this space as splitting up the available phonological phase volume into nearly equidistant volumes. It is shown that this space is ideal for the discussion of such seemingly disparate phenomena as assimilation, metathesis, haplology, and dissimilation. In short this phase space is the natural phase space for speech since it 1 ) provides unification for as diverse phenomena as articulation, acoustics and perception of linguistics; 2 ) explicitly shows how to display spatio-temporal phenomena; 3 ) points the way to improvements using empirical measurements.

Original languageEnglish
Pages (from-to)117-148
Number of pages32
JournalJournal of Quantitative Linguistics
Issue number2
StatePublished - 1999


Dive into the research topics of 'Vector phase space for speech analysis via dimensional analysis'. Together they form a unique fingerprint.

Cite this