Document Type

Student Presentation

Presentation Date


Faculty Sponsor

H Greg Wilson


This research explores the possibility of using a combination of a phonetic feature based binary encoding format for phonemes and dedicated coprocessors to improve computer voice interactions. While both speech synthesis and speech recognition have made great strides recently current performance still leaves much to be desired. Dedicated graphics cards and binary encoding have had a huge impact on computer graphics in the last two decades, and could do the same for voice interactions. Sources for this research consist primarily of course text books and documentation from open source software projects. At the time that this research was conducted it was purely speculative, but since then advances such as Google's TensorFlow AI, and NVIDIA's CUDA development kit make experimental research practical. This research indicates that it would be worthwhile to conduct experimental research on dedicated