VOICE MODIFICATION USING DIGITAL TECHNOLOGY

This report describes the voice modification using digital technology using MATLAB
software and an additional hardware prototype designed. The voice modification
objective that a source voice signal is mapped into another target voice signal. Other
objective of this voice modification is to create a new voice signal from a given source
signal. These modifications are carried by altering the voice waveform features.
By enabling the modification, it is expected that user is able to mimic other person's
voice. In addition, it can be used as a reference and guidance for voice conversion.
For the modification algorithm, an optimization technique for the coding is applied to suit
the objective of the project [7].
The project's major activities are to develop programs using MATLAB, that enable
speech signal recording, analysis, synthesis, modification, and conversion. The recording
is just a simple procedure, where it can be recorded using any computer, with a
microphone. Analysis of speech signal is an essential step where the speech waveform
features are calculated. Here, we model a vocal tract, which resemble as a filter for the
excitation signal input. This filter is designed based on source speech waveform features.
The speech waveform parameters used in this analysis are excitation source model, vocal
tract model and control model which consist of gain and pitch parameter.
Synthesis procedure is a step to produce a synthetic voice. This synthetic voice is just the
display of speech signal designed in the analysis part based on the source signal.
The modification and conversion parts of the software are performed for the voice
modification. The modification part enables the user to change the speech waveform
parameters calculated and thus creates a new voice. The conversion part is actually
performs a mapping of source speech signal to target speech signal.
In this work, hardware also is designed as an additional part to demonstrate the
application. The prototype designed performs the conversion process by modulating the
frequency of input signal.
Tests were carried on both software and hardware. The software conversion design can
convert any source voice signal to target voice signal. But the output will contains some
noises.
The hardware model can also modify any input voice signal to another form [14]. But
modification is very much limited to seven types of output voices. Here also, noises are
noticed.
Since the source codes of the voice conversion software are huge, it is given in Volume 2,
under title Source Code of Voice Modification Software.