Speech and Audio Recognition Linear

Speech is a sequence of small units such as words, vowels and Phonemes. The main objective of this project was synthesise some of these vowels using the source filter model of speech production.

In this project two vowels was selected and then estimated using linear predictive coding LPC.

Figure 1: Male Voice Coeffients

After using LPC then vowels are synthesised by finding their pitch and harmonics. This is done by using Cepstrum.

Figure 2: Male Voice Cepstrum plot

Once they are synthesised they are generated using a periodic impulse train through the pole filter.


Figure 3: Male voice Impulse Train

This assignment was done using Matlab as it contained built in functions for the LPC, filter design, Audio Input and Output.