Studies of Statistical and Syntactic Approaches to Pattern Recognition.
Final scientific rept. 15 Apr 69-31 Jan 74,
PURDUE UNIV LAFAYETTE IND SCHOOL OF ELECTRICAL ENGINEERING
Pagination or Media Count:
Stochastic languages and stochastic syntax analysis for the description and recognition of patterns was investigated. Investigation of syntax-directed recognition techniques applied to certain speech sounds has continued. Visually segmented spectral time samples representing the liquids and glides in isolated utterances have been reduced in dimensionality from 25 dimensions in the original spectral space to two dimensions in each of three recognition spaces. Work has also continued on speaker identification using nasals. Nasal consonants have been demonstrated to be one of the best acoustic cues for speaker identification. Speech data of four speakers were used and showed consistent results. Modified author abstract