Home / Projects / Reaktor Ensembles / LPC Vocoder
LPC Vocoder v1.0

|audio demo|

LPC Vocoder use the "Linear Predictive Coding" that is one of the most powerful speech analysis and synthesis techniques.
It analyzes the speech signal in real-time by estimating the formants, removing their effects from the speech signal, and estimating the intensity and frequency of the remaining sound.
First, speech is sampled at a frequency appropriate, to capture all of the necessary frequency components important for processing and recognition. Once segmented, the speech type is determined as either voiced, unvoiced, or silence. Depending on the V/U/S classification made by the encoder, the excitation source is chose as being either a random noise source or a periodic impulse train.
Formants are determined by autocorrelation method and recursively using the Levinson-Drubin algorithm.

Analysis section contains some important parameters for the voice recognition:
LPC order knob select the order of the LPC From 1 to 10

Input and output level knobs:

There are three leds indicating if a speech segment is voiced, unvoiced or silence

Silence: thresold level to detect silence

Transpose: +24 -24
Midi: play the fundamental frequency by a keyboard


P = inpulse train
D = voice / unvoice decision
N = noise

LPC Vocoder contains many new modules like: autocorrelation, Durbin-Levinson Algorithm, LPC Lattice Implementation Synthesizer, Center Clip, etc.