Realization of Human Voice Synthesis Based on Nyquist

As the field of computer music advances, vocal synthesis has become a prominent frontier of development. Most synthesized results rely on recorded samples as a foundation, either through concatenative synthesis or through training on large voice datasets. On this basis, this research aims to develop a voice synthesizer that does not directly depend on recorded samples, instead using only waveform synthesis. The synthesizer would produce a waveform similar to that of a human larynx, and the result would be passed through several formant filters that make up the oral and glottal cavity. Additional filters and masks would be applied to shape the sounds and produce consonants. The results were semi-successful as the synthesizer was able to produce natural sounding vowels and be modified to have different perceived ages and genders but was unable to form distinguishable consonants. Overall, these results show promise of the possibility of a natural vocal synthesizer without the usage of recorded samples, and further research would be conducted to advance its progress.