A Case Study of Phonetic Reduction in Persian Telephonic Numerals

Phonetic reduction is consistently considered as a challenge for speech recognition systems. Among them is the reduction of Persian telephonic numerals /do-noh/(2-9), /haft-haʃt/ (7-8) and /se-sefr/ (3-0). The acoustic analysis of the telephonic signals /se-sefr/ shows that the deletion of /r/ diminishes the phonological distinctions between [se-sef] pairs, however the phonetic cues in particularly vowel /e/ including vowel duration, F2 and F3 locus equations can still transmit this phonological contrast. Word recognition systems can recover the reduced forms by implementing these acoustic cues.