FreeCalypso > hg > gsm-codec-lib
view libtwamr/d8_31pf.c @ 412:810ac4b99025
libtwamr: integrate VAD2 r_fft.c
author | Mychaela Falconia <falcon@freecalypso.org> |
---|---|
date | Tue, 07 May 2024 01:22:02 +0000 |
parents | 03198f6b0427 |
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 : d8_31pf.c * Purpose : Builds the innovative codevector * ******************************************************************************** */ /* ******************************************************************************** * MODULE INCLUDE FILE AND VERSION ID ******************************************************************************** */ #include "namespace.h" #include "d8_31pf.h" /* ******************************************************************************** * INCLUDE FILES ******************************************************************************** */ #include "typedef.h" #include "basic_op.h" #include "no_count.h" #include "cnst.h" /* ******************************************************************************** * LOCAL VARIABLES AND TABLES ******************************************************************************** */ #define NB_PULSE 8 /* number of pulses */ /* define values/representation for output codevector and sign */ #define POS_CODE 8191 #define NEG_CODE 8191 static void decompress10 ( Word16 MSBs, /* i : MSB part of the index */ Word16 LSBs, /* i : LSB part of the index */ Word16 index1, /* i : index for first pos in pos_index[] */ Word16 index2, /* i : index for second pos in pos_index[] */ Word16 index3, /* i : index for third pos in pos_index[] */ Word16 pos_indx[]) /* o : position of 3 pulses (decompressed) */ { Word16 ia, ib, ic; /* pos_indx[index1] = ((MSBs-25*(MSBs/25))%5)*2 + (LSBs-4*(LSBs/4))%2; pos_indx[index2] = ((MSBs-25*(MSBs/25))/5)*2 + (LSBs-4*(LSBs/4))/2; pos_indx[index3] = (MSBs/25)*2 + LSBs/4; */ test (); if (sub(MSBs, 124) > 0) { MSBs = 124; move16 (); } ia = mult(MSBs, 1311); ia = sub(MSBs, extract_l(L_shr(L_mult(ia, 25), 1))); ib = shl(sub(ia, extract_l(L_shr(L_mult(mult(ia, 6554), 5), 1))), 1); ic = shl(shr(LSBs, 2), 2); ic = sub(LSBs, ic); pos_indx[index1] = add(ib, (ic & 1)); logic16 (); ib = shl(mult(ia, 6554), 1); pos_indx[index2] = add(ib, shr(ic, 1)); pos_indx[index3] = add(shl(mult(MSBs, 1311), 1), shr(LSBs, 2)); return; } /************************************************************************* * * FUNCTION: decompress_code() * * PURPOSE: decompression of the linear codewords to 4+three indeces * one bit from each pulse is made robust to errors by * minimizing the phase shift of a bit error. * 4 signs (one for each track) * i0,i4,i1 => one index (7+3) bits, 3 LSBs more robust * i2,i6,i5 => one index (7+3) bits, 3 LSBs more robust * i3,i7 => one index (5+2) bits, 2-3 LSbs more robust * *************************************************************************/ static void decompress_code ( Word16 indx[], /* i : position and sign of 8 pulses (compressed) */ Word16 sign_indx[], /* o : signs of 4 pulses (signs only) */ Word16 pos_indx[] /* o : position index of 8 pulses (position only) */ ) { Word16 i, ia, ib, MSBs, LSBs, MSBs0_24; for (i = 0; i < NB_TRACK_MR102; i++) { sign_indx[i] = indx[i]; move16 (); } /* First index: 10x10x10 -> 2x5x2x5x2x5-> 125x2x2x2 -> 7+1x3 bits MSBs = indx[NB_TRACK]/8; LSBs = indx[NB_TRACK]%8; */ MSBs = shr(indx[NB_TRACK_MR102], 3); LSBs = indx[NB_TRACK_MR102] & 7; logic16 (); decompress10 (MSBs, LSBs, 0, 4, 1, pos_indx); /* Second index: 10x10x10 -> 2x5x2x5x2x5-> 125x2x2x2 -> 7+1x3 bits MSBs = indx[NB_TRACK+1]/8; LSBs = indx[NB_TRACK+1]%8; */ MSBs = shr(indx[NB_TRACK_MR102+1], 3); LSBs = indx[NB_TRACK_MR102+1] & 7; logic16 (); decompress10 (MSBs, LSBs, 2, 6, 5, pos_indx); /* Third index: 10x10 -> 2x5x2x5-> 25x2x2 -> 5+1x2 bits MSBs = indx[NB_TRACK+2]/4; LSBs = indx[NB_TRACK+2]%4; MSBs0_24 = (MSBs*25+12)/32; if ((MSBs0_24/5)%2==1) pos_indx[3] = (4-(MSBs0_24%5))*2 + LSBs%2; else pos_indx[3] = (MSBs0_24%5)*2 + LSBs%2; pos_indx[7] = (MSBs0_24/5)*2 + LSBs/2; */ MSBs = shr(indx[NB_TRACK_MR102+2], 2); LSBs = indx[NB_TRACK_MR102+2] & 3; logic16 (); MSBs0_24 = shr(add(extract_l(L_shr(L_mult(MSBs, 25), 1)), 12), 5); ia = mult(MSBs0_24, 6554) & 1; ib = sub(MSBs0_24, extract_l(L_shr(L_mult(mult(MSBs0_24, 6554), 5), 1))); test (); if (sub(ia, 1) == 0) { ib = sub(4, ib); } pos_indx[3] = add(shl(ib, 1), (LSBs & 1)); logic16 (); ia = shl(mult(MSBs0_24, 6554), 1); pos_indx[7] = add(ia, shr(LSBs, 1)); } /* ******************************************************************************** * PUBLIC PROGRAM CODE ******************************************************************************** */ /************************************************************************* * * FUNCTION: dec_8i40_31bits() * * PURPOSE: Builds the innovative codevector from the received * index of algebraic codebook. * * See c8_31pf.c for more details about the algebraic codebook structure. * *************************************************************************/ void dec_8i40_31bits ( Word16 index[], /* i : index of 8 pulses (sign+position) */ Word16 cod[] /* o : algebraic (fixed) codebook excitation */ ) { Word16 i, j, pos1, pos2, sign; Word16 linear_signs[NB_TRACK_MR102]; Word16 linear_codewords[NB_PULSE]; for (i = 0; i < L_CODE; i++) { cod[i] = 0; move16 (); } decompress_code (index, linear_signs, linear_codewords); /* decode the positions and signs of pulses and build the codeword */ for (j = 0; j < NB_TRACK_MR102; j++) { /* compute index i */ i = linear_codewords[j]; i = extract_l (L_shr (L_mult (i, 4), 1)); pos1 = add (i, j); /* position of pulse "j" */ test (); if (linear_signs[j] == 0) { sign = POS_CODE; move16 (); /* +1.0 */ } else { sign = -NEG_CODE; move16 (); /* -1.0 */ } cod[pos1] = sign; move16 (); /* compute index i */ i = linear_codewords[add (j, 4)]; i = extract_l (L_shr (L_mult (i, 4), 1)); pos2 = add (i, j); /* position of pulse "j+4" */ test (); if (sub (pos2, pos1) < 0) { sign = negate (sign); } cod[pos2] = add (cod[pos2], sign); move16 (); } return; }