Formant-Based Speech Synthesis: a Case of Amharic Words

No Thumbnail Available



Journal Title

Journal ISSN

Volume Title


Addis Ababa University


Speech synthesis is a process of making artificial speech which mainly requires computers to understand the language speaking rules. Even if there are several techniques of producing synthetic speech, it is still challenging to find one that overcomes all the limitations. One of the speech synthesis techniques is formant synthesis which is based on the well-known source-filter model. This project work has aimed at developing a speech synthesizer for Amharic language words using this technique. The project started by extracting the resonance which is the most important parameter in formant synthesis and other parameters like formant bandwidth, fundamental frequency (pitch), etc from speech file. At the same time the acoustic parameters of the file were passed to the formant synthesizer to synthesize voiced sounds. The unvoiced sounds were segmented from all Amharic syllables and stored in appropriate place. In order to develop the system, there are phases that were performed. These are speech analysis, text analysis and synthesis. In the speech analysis phase, recording of the speech and extraction of the acoustic features from the wave files was made. Segmentation process was conducted before extracting the features of the speech. This process was performed using wavesurfer which is a speech analysis tool used for studies of acoustic phonetics. Therefore, every input text that comes to the system had to be transcribed and produced phonemes from input text. With these phonetic strings, we could generate an artificial speech for the voiced sounds and selecting the consonants with its co-existing context and concatenate these units to synthesize a word. The system provided flexibility of a speech with low memory and data requirements.



A Case of, Amharic Words