A Synthetic Voice
This applet plays two different audio streams. The first is a recording of a man's voice. The second is a synthetic version of the same spoken sentence. The synthetic version is created using an adaptive linear-prediction filter to create an autoregressive model of the speech signal. This filter changes during the speech signal to approximate the spectral characteristics of the speech. These analyzed characteristics are then used to create the synthesized speech.
Play Original Speech
Play Synthesized Speech
The block diagram
for the synthetic speech model.