sarsooxyz.hugo/content/posts/lpss/index.md

960 B

title date draft
Linear Predictive Speech Synthesiser 2020-11-10T14:48:40+00:00 false

During my speech & audio processing & recognition post-grad module, I completed two pieces of coursework. The first of which involved writing and analysing a speech synthesiser utilising linear predictive coding. The report achieved 95%.

spectrogram

Spectrogram analysing one of the original vowel segments; the red circles highlight areas of interest, the lower are the formant frequencies

The report analysed two vowel segments in order to identify their fundamental frequencies and the first handful of formant frequencies. After this, linear predictive coefficients of varying orders were calculated and used in conjunction with the fundamental frequency to re-synthesise the vowel.

GitHub Repo

Read the report here.

lpc