view libtwamr/q_plsf_5.c @ 409:4184ccc136a3

libtwamr/typedef.h: drop Pfloat from EFR, not used in AMR
author Mychaela Falconia <falcon@freecalypso.org>
date Tue, 07 May 2024 01:04:17 +0000
parents 128ec87489b6
children
line wrap: on
line source

/*
********************************************************************************
*
*      GSM AMR-NB speech codec   R98   Version 7.6.0   December 12, 2001
*                                R99   Version 3.3.0                
*                                REL-4 Version 4.1.0                
*
********************************************************************************
*
*      File             : q_plsf_5.c
*      Purpose          : Quantization of 2 sets of LSF parameters using 1st 
*                         order MA prediction and split by 5 matrix
*                         quantization (split-MQ)
*
********************************************************************************
*/
 
/*
********************************************************************************
*                         MODULE INCLUDE FILE AND VERSION ID
********************************************************************************
*/
#include "namespace.h"
#include "q_plsf.h"
 
/*
********************************************************************************
*                         INCLUDE FILES
********************************************************************************
*/
#include "typedef.h"
#include "basic_op.h"
#include "no_count.h"
#include "lsp_lsf.h"
#include "reorder.h"
#include "lsfwt.h"
#include "q_plsf5_tab.h"

/*
********************************************************************************
*                         LOCAL PROGRAM CODE
********************************************************************************
*/
/* Quantization of a 4 dimensional subvector */

static Word16 Vq_subvec (/* o : quantization index,            Q0  */
    Word16 *lsf_r1,      /* i : 1st LSF residual vector,       Q15 */
    Word16 *lsf_r2,      /* i : 2nd LSF residual vector,       Q15 */
    const Word16 *dico,  /* i : quantization codebook,         Q15 */
    Word16 *wf1,         /* i : 1st LSF weighting factors      Q13 */
    Word16 *wf2,         /* i : 2nd LSF weighting factors      Q13 */  
    Word16 dico_size     /* i : size of quantization codebook, Q0  */
)
{
    Word16 index = 0; /* initialization only needed to keep gcc silent */
    Word16 i, temp;
    const Word16 *p_dico;
    Word32 dist_min, dist;

    dist_min = MAX_32;                                  move32 (); 
    p_dico = dico;                                      move16 (); 

    for (i = 0; i < dico_size; i++)
    {
        temp = sub (lsf_r1[0], *p_dico++);
        temp = mult (wf1[0], temp);
        dist = L_mult (temp, temp);

        temp = sub (lsf_r1[1], *p_dico++);
        temp = mult (wf1[1], temp);
        dist = L_mac (dist, temp, temp);

        temp = sub (lsf_r2[0], *p_dico++);
        temp = mult (wf2[0], temp);
        dist = L_mac (dist, temp, temp);

        temp = sub (lsf_r2[1], *p_dico++);
        temp = mult (wf2[1], temp);
        dist = L_mac (dist, temp, temp);

        test (); 
        if (L_sub (dist, dist_min) < (Word32) 0)
        {
            dist_min = dist;                            move32 (); 
            index = i;                                  move16 (); 
        }
    }

    /* Reading the selected vector */

    p_dico = &dico[shl (index, 2)];                     move16 (); 
    lsf_r1[0] = *p_dico++;                              move16 (); 
    lsf_r1[1] = *p_dico++;                              move16 (); 
    lsf_r2[0] = *p_dico++;                              move16 (); 
    lsf_r2[1] = *p_dico++;                              move16 (); 

    return index;
}

/* Quantization of a 4 dimensional subvector with a signed codebook */

static Word16 Vq_subvec_s ( /* o : quantization index            Q0  */
    Word16 *lsf_r1,         /* i : 1st LSF residual vector       Q15 */
    Word16 *lsf_r2,         /* i : and LSF residual vector       Q15 */
    const Word16 *dico,     /* i : quantization codebook         Q15 */
    Word16 *wf1,            /* i : 1st LSF weighting factors     Q13 */
    Word16 *wf2,            /* i : 2nd LSF weighting factors     Q13 */
    Word16 dico_size)       /* i : size of quantization codebook Q0  */  
{
    Word16 index = 0;  /* initialization only needed to keep gcc silent */
    Word16 sign = 0;   /* initialization only needed to keep gcc silent */
    Word16 i, temp;
    const Word16 *p_dico;
    Word32 dist_min, dist;

    dist_min = MAX_32;                                  move32 (); 
    p_dico = dico;                                      move16 (); 

    for (i = 0; i < dico_size; i++)
    {
        /* test positive */

        temp = sub (lsf_r1[0], *p_dico++);
        temp = mult (wf1[0], temp);
        dist = L_mult (temp, temp);

        temp = sub (lsf_r1[1], *p_dico++);
        temp = mult (wf1[1], temp);
        dist = L_mac (dist, temp, temp);

        temp = sub (lsf_r2[0], *p_dico++);
        temp = mult (wf2[0], temp);
        dist = L_mac (dist, temp, temp);

        temp = sub (lsf_r2[1], *p_dico++);
        temp = mult (wf2[1], temp);
        dist = L_mac (dist, temp, temp);

        test (); 
        if (L_sub (dist, dist_min) < (Word32) 0)
        {
            dist_min = dist;                            move32 (); 
            index = i;                                  move16 (); 
            sign = 0;                                   move16 (); 
        }
        /* test negative */

        p_dico -= 4;                                    move16 (); 
        temp = add (lsf_r1[0], *p_dico++);
        temp = mult (wf1[0], temp);
        dist = L_mult (temp, temp);

        temp = add (lsf_r1[1], *p_dico++);
        temp = mult (wf1[1], temp);
        dist = L_mac (dist, temp, temp);

        temp = add (lsf_r2[0], *p_dico++);
        temp = mult (wf2[0], temp);
        dist = L_mac (dist, temp, temp);

        temp = add (lsf_r2[1], *p_dico++);
        temp = mult (wf2[1], temp);
        dist = L_mac (dist, temp, temp);

        test (); 
        if (L_sub (dist, dist_min) < (Word32) 0)
        {
            dist_min = dist;                            move32 (); 
            index = i;                                  move16 (); 
            sign = 1;                                   move16 (); 
        }
    }

    /* Reading the selected vector */

    p_dico = &dico[shl (index, 2)];                     move16 (); 
    test (); 
    if (sign == 0)
    {
        lsf_r1[0] = *p_dico++;                          move16 (); 
        lsf_r1[1] = *p_dico++;                          move16 (); 
        lsf_r2[0] = *p_dico++;                          move16 (); 
        lsf_r2[1] = *p_dico++;                          move16 (); 
    }
    else
    {
        lsf_r1[0] = negate (*p_dico++);                 move16 (); 
        lsf_r1[1] = negate (*p_dico++);                 move16 (); 
        lsf_r2[0] = negate (*p_dico++);                 move16 (); 
        lsf_r2[1] = negate (*p_dico++);                 move16 (); 
    }

    index = shl (index, 1);
    index = add (index, sign);

    return index;
}

/*
********************************************************************************
*                         PUBLIC PROGRAM CODE
********************************************************************************
*/
 
/*************************************************************************
 *   FUNCTION:  Q_plsf_5()
 *
 *   PURPOSE:  Quantization of 2 sets of LSF parameters using 1st order MA
 *             prediction and split by 5 matrix quantization (split-MQ)
 *
 *   DESCRIPTION:
 *
 *        p[i] = pred_factor*past_rq[i];   i=0,...,m-1
 *        r1[i]= lsf1[i] - p[i];           i=0,...,m-1
 *        r2[i]= lsf2[i] - p[i];           i=0,...,m-1
 *   where:
 *        lsf1[i]           1st mean-removed LSF vector.
 *        lsf2[i]           2nd mean-removed LSF vector.
 *        r1[i]             1st residual prediction vector.
 *        r2[i]             2nd residual prediction vector.
 *        past_r2q[i]       Past quantized residual (2nd vector).
 *
 *   The residual vectors r1[i] and r2[i] are jointly quantized using
 *   split-MQ with 5 codebooks. Each 4th dimension submatrix contains 2
 *   elements from each residual vector. The 5 submatrices are as follows:
 *     {r1[0], r1[1], r2[0], r2[1]};  {r1[2], r1[3], r2[2], r2[3]};
 *     {r1[4], r1[5], r2[4], r2[5]};  {r1[6], r1[7], r2[6], r2[7]};
 *                    {r1[8], r1[9], r2[8], r2[9]};
 *
 *************************************************************************/
void Q_plsf_5 (
    Q_plsfState *st,
    Word16 *lsp1,      /* i : 1st LSP vector,                     Q15 */
    Word16 *lsp2,      /* i : 2nd LSP vector,                     Q15 */   
    Word16 *lsp1_q,    /* o : quantized 1st LSP vector,           Q15 */
    Word16 *lsp2_q,    /* o : quantized 2nd LSP vector,           Q15 */
    Word16 *indice     /* o : quantization indices of 5 matrices, Q0  */
)
{
    Word16 i;
    Word16 lsf1[M], lsf2[M], wf1[M], wf2[M], lsf_p[M], lsf_r1[M], lsf_r2[M];
    Word16 lsf1_q[M], lsf2_q[M];

    /* convert LSFs to normalize frequency domain 0..16384  */

    Lsp_lsf (lsp1, lsf1, M);
    Lsp_lsf (lsp2, lsf2, M);

    /* Compute LSF weighting factors (Q13) */
    
    Lsf_wt (lsf1, wf1);
    Lsf_wt (lsf2, wf2);

    /* Compute predicted LSF and prediction error */

    for (i = 0; i < M; i++)
    {
        lsf_p[i] = add (mean_lsf[i], mult (st->past_rq[i], LSP_PRED_FAC_MR122));
        move16 (); 
        lsf_r1[i] = sub (lsf1[i], lsf_p[i]);           move16 (); 
        lsf_r2[i] = sub (lsf2[i], lsf_p[i]);           move16 (); 
    }

    /*---- Split-MQ of prediction error ----*/

    indice[0] = Vq_subvec (&lsf_r1[0], &lsf_r2[0], dico1_lsf,
                           &wf1[0], &wf2[0], DICO1_SIZE);
                                                        move16 (); 

    indice[1] = Vq_subvec (&lsf_r1[2], &lsf_r2[2], dico2_lsf,
                           &wf1[2], &wf2[2], DICO2_SIZE);
                                                        move16 (); 

    indice[2] = Vq_subvec_s (&lsf_r1[4], &lsf_r2[4], dico3_lsf,
                             &wf1[4], &wf2[4], DICO3_SIZE);
                                                        move16 (); 

    indice[3] = Vq_subvec (&lsf_r1[6], &lsf_r2[6], dico4_lsf,
                           &wf1[6], &wf2[6], DICO4_SIZE);
                                                        move16 (); 

    indice[4] = Vq_subvec (&lsf_r1[8], &lsf_r2[8], dico5_lsf,
                           &wf1[8], &wf2[8], DICO5_SIZE);
                                                        move16 (); 

    /* Compute quantized LSFs and update the past quantized residual */
    for (i = 0; i < M; i++)
    {
        lsf1_q[i] = add (lsf_r1[i], lsf_p[i]);          move16 (); 
        lsf2_q[i] = add (lsf_r2[i], lsf_p[i]);          move16 (); 
        st->past_rq[i] = lsf_r2[i];                     move16 (); 
    }

    /* verification that LSFs has minimum distance of LSF_GAP */

    Reorder_lsf (lsf1_q, LSF_GAP, M);
    Reorder_lsf (lsf2_q, LSF_GAP, M);

    /*  convert LSFs to the cosine domain */
    
    Lsf_lsp (lsf1_q, lsp1_q, M);
    Lsf_lsp (lsf2_q, lsp2_q, M);
}