KSPAN

NAME

K-SPAN - programming language for Klatt synthesizer control

DESCRIPTION

K-SPAN is a programming language which can be used to construct sets of control parameters for producing synthetic speech with the Klatt formant synthesizer. You should view a K-SPAN program as a specification for constructing a number of synthesizer control parameter tables, with each parameter specified every 5ms. The advantage of a K-SPAN program over the parameter table is that interpolations within the table are calculated automatically, so that only turning points in a parameter specification with time need to be specified. In addition, the K-SPAN language allows for the construction of a number of syntheses sharing common components, and a K-SPAN program may be readily constructed for building a set of stimuli for use in a speech perception experiment, for example.

STRUCTURE

K-SPAN is actually implemented as a set of macros for a standard C-language compiler. When a K-SPAN program is compiled, it is built into a C source module and linked with a standard run-time object module to construct a program which when run builds one or more synthesizer parameter tables from the K-SPAN program specification. These tables are saved in files suitable from input to the Klatt software synthesizer (see klattsyn(1) ) For details of compilation and execution of K-SPAN programs see the KSPAN manual page.

The benefit of the C-language implementation is that K-SPAN programs can make use of all the facilities of the C language in addition to the commands provided by K-SPAN which are detailed below. Thus a K-SPAN program can use global and local variables, call sub-routines and use the C pre-processor to modify the command language.

"SOURCE LANGUAGE CONVENTIONS"

The basic source file structure follows the C language. Statements must be separated by semicolons, Comments must be enclosed in /* ... */. The main subroutine must be called 
void SYNTH(char **param)
The 'param' arguments are character strings from the command line from switches -A <value>, -B <value>, etc. The user is free to construct other subroutines which can be called from SYNTH(). Thus the basic C-SPAN program looks like: 

/* comments */
void SYNTH(char **param)
{
	/* parameter settings in here */
}

"SYNTHESIS CONTROL COMMANDS"

The following commands are available for overall control of a synthesis:

LENGTH(t) This sets the overall length of the synthesis to t milliseconds. The default length is taken from the lastest time at which an assignment is made. You need this command if modifications to a synthesis reduce its length.

CLEAR This clears the entire synthesis table, and resets the start time and end times to zero.

FLUSH This command forces construction of the synthesizer parameter table at this point in the specification. The synthesizer parameter table is constructed automatically at the end of the SYNTH() routine, so you only need this command if your program builds more than one synthesis.

SAVE This command saves the current time position in the synthesis on an internal stack. Since all settings of AT(t) and WAIT(t) affect the current time position, regardless of whether they are made in SYNTH() or in a subroutine, you may need to save the current time to make local parameter settings and then restore it.

RESTORE This command restores the last current time position save with SAVE.

"PARAMETER SETTINGS"

The parameter setting statements are all of the form <paramname>(<value>,<interpolation>) where <paramname> is from the list:

F0(v,i) Fundamental frequency, in Hz. Range: 0 to 500. Default 100. (NOTE FX() may also be used)

AV(v,i) Amplitude of voicing, in dB. Range: 0 to 80. Default 60.

OQ(v,i) Open quotient (voicing open-time/period), in %. Range: 10 to 99. Default 50.

SQ(v,i) Speed quotient (rise/fall time, LF model), in %. Range: 100 to 500. Default 200.

TL(v,i) Extra tilt of voicing spectrum, dB down @ 3 kHz. Range: 0 to 41. Default 0.

FL(v,i) Flutter (random fluct in f0), in % of maximum. Range: 0 to 100. Default 0.

DI(v,i) Diplophonia (alt periods closer), in % of max. Range: 0 to 100. Default 0.

AH(v,i) Amplitude of aspiration, in dB. Range: 0 to 80. Default 0.

AF(v,i) Amplitude of frication, in dB. Range: 0 to 80. Default 0.

F1(v,i) Frequency of 1st formant, in Hz. Range: 180 to 1300. Default 500.

B1(v,i) Bandwidth of 1st formant, in Hz. Range: 30 to 1000. Default 60.

DF1(v,i) Change in F1 during open portion of period, in Hz. Range: 0 to 100. Default 0.

DB1(v,i) Change in B1 during open portion of period, in Hz. Range: 0 to 400. Default 0.

F2(v,i) Frequency of 2nd formant, in Hz. Range: 550 to 3000. Default 1500.

B2(v,i) Bandwidth of 2nd formant, in Hz. Range: 40 to 1000. Default 90.

F3(v,i) Frequency of 3rd formant, in Hz. Range: 1200 to 4800. Default 2500.

B3(v,i) Bandwidth of 3rd formant, in Hz. Range: 60 to 1000. Default 150.

F4(v,i) Frequency of 4th formant, in Hz. Range: 2400 to 4990. Default 3250.

B4(v,i) Bandwidth of 4th formant, in Hz. Range: 100 to 1000. Default 200.

F5(v,i) Frequency of 5th formant, in Hz. Range: 3000 to 4990. Default 3700.

B5(v,i) Bandwidth of 5th formant, in Hz. Range: 100 to 1500. Default 200.

F6(v,i) Frequency of 6th formant, in Hz (applies if NF=6). Range: 3000 to 4990. Default 4990.

B6(v,i) Bandwidth of 6th formant, in Hz (applies if NF=6). Range: 100 to 4000. Default 500.

FNP(v,i) Frequency of nasal pole, in Hz. Range: 180 to 500. Default 280.

BNP(v,i) Bandwidth of nasal pole, in Hz. Range: 40 to 1000. Default 90.

FNZ(v,i) Frequency of nasal zero, in Hz. Range: 180 to 800. Default 280.

BNZ(v,i) Bandwidth of nasal zero, in Hz. Range: 40 to 1000. Default 90.

FTP(v,i) Frequency of tracheal pole, in Hz. Range: 300 to 3000. Default 2150.

BTP(v,i) Bandwidth of tracheal pole, in Hz. Range: 40 to 1000. Default 180.

FTZ(v,i) Frequency of tracheal zero, in Hz. Range: 300 to 3000. Default 2150.

BTZ(v,i) Bandwidth of tracheal zero, in Hz. Range: 40 to 2000. Default 180.

A2F(v,i) Amplitude of fric-excited parallel 2nd formant, in dB. Range: 0 to 80. Default 0.

A3F(v,i) Amplitude of fric-excited parallel 3rd formant, in dB. Range: 0 to 80. Default 0.

A4F(v,i) Amplitude of fric-excited parallel 4th formant, in dB. Range: 0 to 80. Default 0.

A5F(v,i) Amplitude of fric-excited parallel 5th formant, in dB. Range: 0 to 80. Default 0.

A6F(v,i) Amplitude of fric-excited parallel 6th formant, in dB. Range: 0 to 80. Default 0.

AB(v,i) Amplitude of fric-excited parallel bypass path, in dB. Range: 0 to 80. Default 0.

B2F(v,i) Bandwidth of fric-excited parallel 2nd formant, in Hz. Range: 40 to 1000. Default 250.

B3F(v,i) Bandwidth of fric-excited parallel 3rd formant, in Hz. Range: 60 to 1000. Default 300.

B4F(v,i) Bandwidth of fric-excited parallel 4th formant, in Hz. Range: 100 to 1000. Default 320.

B5F(v,i) Bandwidth of fric-excited parallel 5th formant, in Hz. Range: 100 to 1500. Default 360.

B6F(v,i) Bandwidth of fric-excited parallel 6th formant, in Hz. Range: 100 to 4000. Default 1500.

ANV(v,i) Amplitude of voice-excited parallel nasal formant, in dB. Range: 0 to 80. Default 0.

A1V(v,i) Amplitude of voice-excited parallel 1st formant, in dB. Range: 0 to 80. Default 60.

A2V(v,i) Amplitude of voice-excited parallel 2nd formant, in dB. Range: 0 to 80. Default 60.

A3V(v,i) Amplitude of voice-excited parallel 3rd formant, in dB. Range: 0 to 80. Default 60.

A4V(v,i) Amplitude of voice-excited parallel 4th formant, in dB. Range: 0 to 80. Default 60.

ATV(v,i) Amplitude of voice-excited par tracheal formant, in dB. Range: 0 to 80. Default 0.

and where <value> is a decimal number representing the value to which to set the control parameter at the current time. The value NUL can be used to signify that no change to the currently specified value at this position is requested (if only the interpolation type was to be changed, for example).

The parameter <interpolation> is from this list:

FIX Set value at this time with no interpolation.

LIN Set value at this time and interpolate to the first subsequent specification for this parameter. Interpolations are actually performed after all specifications are made, so you do not need to specify the end value of the interpolation before the start value.

LOG Set value at this time and logarithmically interpolate to the first subsequent setting of this parameter.

NUL Make no changes to the interpolation setting currently specified at this time position.

In addition to these parameter settings the command GET(p,t) returns the value of parameter p (FX,F1, etc) at time t. The global variable NOW contains the current time position. This command does not make any changes to the parameter table. If you request the value of a parameter in an interpolated region, GET returns the interpolated value using the current specifications for the start and end points.

There is no range checking on the values used in these parameter settings in a K-SPAN program.

The file kspan.h contains some useful macros to simplify the command language by using linear interpolations everywhere by default.

"TIMING COMMANDS"

AT(t) Set the current time position in the synthesis to absolute t milliseconds. Only one current time position is maintained globally, so be careful of constructions such as: 


AT(100);
rampamplitudes();
FX(90,FIX);		/* does NOT change FX at time 100 */
 :
 :
rampamplitudes()
{
	A1(NUL,LIN); A2(NUL,LIN); A3(NUL,LIN);
	WAIT(50);
	A1(0,FIX); A2(0,FIX); A3(0,FIX);
}
Use the SAVE and RESTORE commands within subroutines to preserve the current time position outside the routine.

WAIT(t) Move the current time position on by t milliseconds.

The global variable NOW contains the current time position.

"EXAMPLE PROGRAM"


/* test.spk - demonstration K-SPAN program */

#define SILENCE { 
		AV(0); AH(0); AF(0); 
		A2F(0); A3F(0); A4F(0); A5F(0); A6F(0); AB(0); 
		A1V(0); A2V(0); A3V(0); A4V(0); ATV(0); 
		}

void SYNTH(char **param)
{
/* Synthesis 0 */
AT(0);
	/* set up central vowel */
	FX(120,LOG);
	F1(500,LIN); F2(1500,LIN); F3(2500,LIN);
	SILENCE;
	AV(60,FIX);

WAIT(50);
	/* ramp up amplitudes */
	A1V(40,FIX); A2V(35,FIX); A3V(30,FIX);

AT(500);
	/* ramp down amplitudes */
	A1V(NUL,LIN); A2V(NUL,LIN); A3V(NUL,LIN);

WAIT(50);
	/* end at /i/ vowel */
	FX(90,FIX);
	F1(250,FIX); F2(2500,FIX); F3(3300,FIX);
	SILENCE;

	/* make this synthesis */
	FLUSH;

/* Synthesis 1 */
AT(550);
	/* change end to /a/ vowel */
	F1(600,FIX);	F2(750,FIX);	F3(2500,FIX);

	/* and re-make it */
	FLUSH;

}

VERSION/AUTHOR

1.0 Mark Huckvale

SEE ALSO

kspan(1), c-span(5), enspan(1), klattsyn(1)

BUGS

Should be a mechanism for using parameter names as variables.
Fri Jul 09 14:54:50 2004