Skip to content

crispinprojects/diphone-talker

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

9 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Diphone Speech Synthesizer

A code example of a diphone speech synthesizer for speaking dates. It is some of the early development code for the Talk Calendar project. It has been developed using C and is a Linux console application.

How is Speech Generated?

Words are formed as sequences of elementary speech units. A phoneme is the smallest unit of sound that distinguishes one word from another word and there are 44 phonemes in the English language. A diphone is a sound unit composed of two adjacent partial phonemes i.e. the second half of the first phoneme and the first half of the second phoneme. The synthesizer uses a set of pre-recorded diphone sound samples and concatenates diphone audio to produce speech output for a given text input.

If phoneme one is represented as p1=[1 2 3] and phoneme two p2=[4 5 6] then the diphone derived from p1 and p2 is d1-2 =[2 3 4 5]. Phonemes are the "steady state" of a sound while diphones capture coarticulation which is the way the mouth moves from the middle of one sound to the middle of the next. The most complex part of speech is the transition between sounds. Recording the transition (i.e. the diphone) and then "gluing" diphones together at the stable centres of the phonemes produces much more natural speech than just playing phonemes back-to-back. With 44 phonemes there would theoretically be 44 squared diphone combinations (=1936). However, many combinations never occur in natural speech or are very rare and so typically a diphone count of around 1600 is used with standard English.

The voice used by the diphone speech synthesizer is derivative work based on the diphone collection created by Alan W Black and Kevin Lenzo which is free for use for any purpose (commercial or otherwise) and subject to the light restrictions detailed here. There is information about recording your own diphones here and in the speech synthesis lecture by Professor Alan W Black here.

Build From Source

The C source code for the diphone speech synthesizer for reading out dates is provided in the src directory and has been developed using Debian Trixie.

Building on Debian Trixie

To build the diphone speech synthesizer from source you need the gcc compiler and the Make build system. You need to install the following package.

sudo apt install build-essential

The package:

sudo apt install alsa-utils

is needed but should be installed by default.

Use the MAKEFILE to compile the code. Just run "make" inside the source code folder.

make

To run the diphone speech synthesizer called talker from the terminal use

./talker

This will read out the current date.

Summary

The code has been developed from scratch and is not dependent on any other audio processing libraries and provides a working C code example of a diphone speech synthesizer which may be useful for educational purposes. It compiles and runs and reads out a date. I have developed a talking calendar called Talk Calendar which uses a more advanced diphone concatenation speech synthesizer.

Versioning

SemVer is used for versioning. The version number has the form 0.0.0 representing major, minor and bug fix changes.

Author

Project Status

Active.

Acknowledgements

  • Diphone Source and License

  • Diphone collection and synthesis Alan W. Black and Kevin Lenzo 2000

  • Geany can be used as a source code editor. Geany is lightweight and has an integrated terminal for building the application. To install the Geany IDE use the command below.

  • Talk Calendar

About

Diphone Speech Synthesizer for speaking dates

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages