pitsyn.f,v

这是LPC-10压缩算法的源代码,愿共享之.也希望能赐与MELP方面的算法源码.

head	1.2;
access;
symbols;
locks; strict;
comment	@* @;


1.2
date	96.03.25.18.49.07;	author jaf;	state Exp;
branches;
next	1.1;

1.1
date	96.02.07.14.48.18;	author jaf;	state Exp;
branches;
next	;


desc
@@


1.2
log
@Added commments about which indices of array arguments are read or
written.

Rearranged local variable declarations to indicate which need to be
saved from one invocation to the next.  Added entry INITPITSYN to
reinitialize local state variables, if desired.

Added lots of comments about proving that the maximum number of pitch
periods (NOUT) that can be returned is 16.  The call to STOP that
could happen if NOUT got too large was removed as a result.

Also proved that the total number of samples returned from N calls,
each with identical values of LFRAME, will always be in the range
N*LFRAME-MAXPIT+1 to N*LFRAME.
@
text
@******************************************************************
*
*	PITSYN Version 53
*
* $Log: pitsyn.f,v $
* Revision 1.1  1996/02/07 14:48:18  jaf
* Initial revision
*
*
******************************************************************
*
*   Synthesize a single pitch epoch
*
* Input:
*  ORDER  - Synthesis order (number of RC's)
*  VOICE  - Half frame voicing decisions
*           Indices 1 through 2 read.
*  LFRAME - Length of speech buffer
* Input/Output:
*  PITCH  - Pitch
*           This value should be in the range MINPIT (20) to MAXPIT
*           (156), inclusive.
*           PITCH can be modified under some conditions.
*  RMS    - Energy  (can be modified)
*           RMS is changed to 1 if the value passed in is less than 1.
*  RC     - Reflection coefficients
*           Indices 1 through ORDER can be temporarily overwritten with
*           RCO, and then replaced with original values, under some
*           conditions.
* Output:
*  IVUV   - Pitch epoch voicing decisions
*           Indices (I) of IVUV, IPITI, and RMSI are written,
*           and indices (J,I) of RCI are written,
*           where I ranges from 1 to NOUT, and J ranges from 1 to ORDER.
*  IPITI  - Pitch epoch length
*  RMSI   - Pitch epoch energy
*  RCI    - Pitch epoch RC's
*  NOUT   - Number of pitch periods in this frame
*           This is at least 0, at least 1 if MAXPIT .LT. LFRAME (this
*           is currently true on every call), and can never be more than
*           (LFRAME+MAXPIT-1)/PITCH, which is currently 16 with
*           LFRAME=180, MAXPIT=156, and PITCH .GE. 20, as SYNTHS
*           guarantees when it calls this subroutine.
*  RATIO  - Previous to present energy ratio
*           Always assigned a value.
*  
	SUBROUTINE PITSYN( ORDER, VOICE, PITCH, RMS, RC, LFRAME,
     1  		   IVUV, IPITI, RMSI, RCI, NOUT, RATIO )
	INCLUDE 'config.fh'

*       Arguments

	INTEGER ORDER, VOICE(2), PITCH
	REAL RMS, RC(ORDER)
	INTEGER LFRAME
	INTEGER IVUV(16), IPITI(16)
	REAL RMSI(16), RCI(ORDER,16)
	INTEGER NOUT
	REAL RATIO

*       Local variables that need not be saved

*       LSAMP is initialized in the IF (FIRST) THEN clause, but it is
*       not used the first time through, and it is given a value before
*       use whenever FIRST is .FALSE., so it appears unnecessary to
*       assign it a value when FIRST is .TRUE.

	INTEGER I, J, LSAMP, IP, ISTART, IVOICE
	INTEGER JUSED, NL
	REAL ALRN, ALRO, PROP
	REAL SLOPE, UVPIT, XXY
	INTEGER VFLAG
	REAL YARC(MAXORD)

*       Local state

* FIRST  - .TRUE. only on first call to PITSYN.
* IVOICO - Previous VOICE(2) value.
* IPITO  - Previous PITCH value.
* RMSO   - Previous RMS value.
* RCO    - Previous RC values.
* 
* JSAMP  - If this routine is called N times with identical values of
*          LFRAME, then the total length of all pitch periods returned
*          is always N*LFRAME-JSAMP, and JSAMP is always in the range 0
*          to MAXPIT-1 (see below for why this is so).  Thus JSAMP is
*          the number of samples "left over" from the previous call to
*          PITSYN, that haven't been "used" in a pitch period returned
*          from this subroutine.  Every time this subroutine is called,
*          it returns pitch periods with a total length of at most
*          LFRAME+JSAMP.
*          
* IVOICO, IPITO, RCO, and JSAMP need not be assigned an initial value
* with a DATA statement, because they are always initialized on the
* first call to PITSYN.
*          
* FIRST and RMSO should be initialized with DATA statements, because
* even on the first call, they are used before being initialized.

	INTEGER IVOICO, IPITO
	REAL RMSO
	REAL RCO(MAXORD)
	INTEGER JSAMP
	LOGICAL FIRST

	SAVE IVOICO, IPITO, RMSO, RCO, JSAMP, FIRST

	DATA RMSO /1./
	DATA FIRST /.TRUE./


	IF (RMS .LT. 1) RMS = 1
	IF (RMSO .LT. 1) RMSO = 1
	UVPIT = 0.0
	RATIO = RMS/(RMSO+8.)
	IF (FIRST) THEN
	   LSAMP = 0
	   IVOICE = VOICE(2)
	   IF (IVOICE .EQ. 0) PITCH = LFRAME/4
	   NOUT = LFRAME/PITCH
	   JSAMP = LFRAME - NOUT*PITCH
*          
*          SYNTHS only calls this subroutine with PITCH in the range 20
*          to 156.  LFRAME = MAXFRM = 180, so NOUT is somewhere in the
*          range 1 to 9.
*          
*          JSAMP is "LFRAME mod PITCH", so it is in the range 0 to
*          (PITCH-1), or 0 to MAXPIT-1=155, after the first call.
*          
	   DO I = 1,NOUT
	      DO J = 1,ORDER
	         RCI(J,I) = RC(J)
	      END DO
	      IVUV(I) = IVOICE
	      IPITI(I) = PITCH
	      RMSI(I) = RMS
	   END DO
	   FIRST = .FALSE.
	ELSE
	   VFLAG = 0
	   LSAMP = LFRAME + JSAMP
	   SLOPE = (PITCH-IPITO)/FLOAT(LSAMP)
	   NOUT = 0
	   JUSED = 0
	   ISTART = 1
	   IF ((VOICE(1) .EQ. IVOICO) .AND. (VOICE(2) .EQ. VOICE(1))) THEN
	      IF (VOICE(2) .EQ. 0) THEN
* SSUV - -   0  ,  0  ,  0
	         PITCH = LFRAME/4
	         IPITO = PITCH
	         IF (RATIO .GT. 8) RMSO = RMS
	      END IF
* SSVC - -   1  ,  1  ,  1
	      SLOPE = (PITCH-IPITO)/FLOAT(LSAMP)
	      IVOICE = VOICE(2)
	   ELSE
	      IF (IVOICO .NE. 1) THEN
	         IF (IVOICO .EQ. VOICE(1)) THEN
* UV2VC2 - -  0  ,  0  ,  1
	            NL = LSAMP - LFRAME/4
	         ELSE
* UV2VC1 - -  0  ,  1  ,  1
	            NL = LSAMP - 3*LFRAME/4
	         ENDIF
	         IPITI(1) = NL/2
	         IPITI(2) = NL - IPITI(1)
	         IVUV(1) = 0
	         IVUV(2) = 0
	         RMSI(1) = RMSO
	         RMSI(2) = RMSO
	         DO I = 1,ORDER
	            RCI(I,1) = RCO(I)
	            RCI(I,2) = RCO(I)
	            RCO(I)   = RC(I)
	         END DO
	         SLOPE = 0
	         NOUT = 2
	         IPITO = PITCH
	         JUSED = NL
	         ISTART = NL + 1
	         IVOICE = 1
	      ELSE
	         IF (IVOICO .NE. VOICE(1)) THEN
* VC2UV1 - -   1  ,  0  ,  0
	            LSAMP = LFRAME/4 + JSAMP
	         ELSE
* VC2UV2 - -   1  ,  1  ,  0
	            LSAMP = 3*LFRAME/4 + JSAMP
	         END IF
	         DO I = 1,ORDER
	            YARC(I) = RC(I)
	            RC(I) = RCO(I)
	         END DO
	         IVOICE = 1
	         SLOPE = 0.
	         VFLAG = 1
	      END IF
	   END IF

* Here is the value of most variables that are used below, depending on
* the values of IVOICO, VOICE(1), and VOICE(2).  VOICE(1) and VOICE(2)
* are input arguments, and IVOICO is the value of VOICE(2) on the
* previous call (see notes for the IF (NOUT .NE. 0) statement near the
* end).  Each of these three values is either 0 or 1.  These three
* values below are given as 3-bit long strings, in the order IVOICO,
* VOICE(1), and VOICE(2).  It appears that the code above assumes that
* the bit sequences 010 and 101 never occur, but I wonder whether a
* large enough number of bit errors in the channel could cause such a
* thing to happen, and if so, could that cause NOUT to ever go over 11?
* 
* Note that all of the 180 values in the table are really LFRAME, but
* 180 has fewer characters, and it makes the table a little more