A Novel FPGA Based Low Cost Solution for Tamil-text to Speech Synthesizer

摘要

This study presents a prior work of developing a single chip solution for Text-to-Speech synthesizer for Tamil (Tamil-TTS) language. Though there are enormous works presented in the recent days to address TTS for their native languages, the motivation of this study is to develop a low-cost FPGA based solution for Tamil TTS synthesizer. This study uses the unique feature of Tamil language to eliminate the complexity involved in accessing a database of stored audio signals. It uses only the audio signals of consonants and vowels in the stored memory locations. The compound characters from the segmented input text are generated using a Direct Digital Synthesizer by operating at three different frequencies of phonetic interval units of Tamil. The proposed system is implemented in Cyclone IVE EP4CE115F29C7 FPGA device and the implementation results show that the proposed system outperforms the other similar methods in terms of memory utilization, text-to-speech time, area utilization and power dissipation. The accuracy of the system is examined with 25 native speakers and acceptable accuracy scale has been reached.