The Data

Contents Home The Problem The ARMA Model Autocorrelation What we did [The Data] Conclusions A Word on Decoding Acknowledgements Who we are

In the graphs below, there is a noticeable reduction in amplitude between the original signal and the error. The input signals are the spoken vowels: “aaah”    female voice, "aaah" sound(source .wav file) “oooh”   Male voice, "oooh" sound(source .wav file) Since the error decreases the amplitude by 50-75%, it takes 1 to 2 bits less per sample to transmit with the same resolution. This error is necessary for correct reconstruction because the prediction does not take into account all of the minor variations between periods.

Copyright (c) 2000 by the Oracle Gang.