Amharic speech synthesis using cepstral method with stress generation rule

Amharic is the official language of Ethiopia. In this paper, we present our study on Amharic stress. Stress (Gemination of consonants) in Amharic language is very important for proper pronunciation of words. It is also one of the most distinctive characteristics of the rhythm of the speech. We discuss a method employed for generating stressed syllables from unstressed syllables, and its application to our speech synthesizer. First, we analyzed waveforms of minimal pair words concerned with stressed and unstressed syllables into the time patterns of pitch, power and spectrum. Then, by combining or exchanging these patterns, speech sounds were synthesized. Using the synthesized sounds, listening tests were performed to examine the acoustic correlates of stress among pitch, spectrum, power and duration. We found that consonant’s duration is the most important factor. A further listening test was performed to determine the threshold of duration of consonants between unstressed and stressed syllables, and we observed that 50ms is the average threshold duration for voiced consonants and 70ms is for unvoiced consonants.