phosynth - formant synthesis by rule from MBROLA format control file


phosynth (-I) (-l|-s) (-m file.pho|-t transcription) (-a) (-p patchfile) (out.sfs)


phosynth is a program to perform synthesis by rule for a formant synthesizer. It takes as input a phonetic transcription and a fundamental frequency contour and produces synthesizer control data. The input can come from an input AN item and FX item or from a .PHO file as used by the MBROLA diphone synthesis system. Output is direct to the synthesizer, or to a listing file or to an SFS file.

The formant parameter tracks are produced using the Holmes, Mattingly and Shearme algorithm (1964) and target values are adapted from the JSRU text to speech system.

Here is an example for you to try:

_ 50
t 95 0 150
e 133 0 150 99 110
s 150
t 135 0 110
_ 50


-I Identify program name and version number.

-i item Select input AN and FX items.

-l Send synthesizer control data to standard output in format compatible with sylist(SFS1) and syload(SFS1).

-s Save synthesizer control data to SFS file.

-m mbrola.pho Take input from MBROLA format .PHO file.

-t transcription Take input from a phonetic transcription supplied on command line. For example "_ t e s t _".

-a Save transcription as annotations to SFS file specified (must exist). For transcription (-t) input only.

-p patchfile Reads in patches to the internal segment parameter table from the supplied text file. Commands in the file are separated by white space, commas or semi-colons. Each command is of the form <parameter>=<value>. Available parameters are: segment, rank, target<PAR>, externdur<PAR>, interndur<PAR>. The PAR values are; FN, F1, F2, F3, AL, A1, A2, A3, AH, VR. The segment command sets the context for the subsequent commands until the next segment command. For example:

segment=A:, targetF1=700, targetF2=1060, targetF3=2950
Lines starting with '#' are considered comments.


1.4 Mark Huckvale




Fri Jul 09 14:54:43 2004