FreeCalypso > hg > gsm-codec-lib
view libgsmefr/lsp_lsf.c @ 530:96c4ed5529bf
libgsmfr2 preproc: implement support for DTXd
author | Mychaela Falconia <falcon@freecalypso.org> |
---|---|
date | Thu, 19 Sep 2024 20:15:54 +0000 |
parents | bcd337afa0f1 |
children |
line wrap: on
line source
/************************************************************************* * * FUNCTIONS: Lsp_lsf and Lsf_lsp * * PURPOSE: * Lsp_lsf: Transformation lsp to lsf * Lsf_lsp: Transformation lsf to lsp * * DESCRIPTION: * lsp[i] = cos(2*pi*lsf[i]) and lsf[i] = arccos(lsp[i])/(2*pi) * * The transformation from lsp[i] to lsf[i] and lsf[i] to lsp[i] are * approximated by a look-up table and interpolation. * *************************************************************************/ #include "gsm_efr.h" #include "typedef.h" #include "namespace.h" #include "basic_op.h" #include "no_count.h" #include "sig_proc.h" #include "lsp_lsf.tab" /* Look-up table for transformations */ void Lsf_lsp ( Word16 lsf[], /* (i) : lsf[m] normalized (range: 0.0<=val<=0.5) */ Word16 lsp[], /* (o) : lsp[m] (range: -1<=val<1) */ Word16 m /* (i) : LPC order */ ) { Word16 i, ind, offset; Word32 L_tmp; for (i = 0; i < m; i++) { ind = shr (lsf[i], 8); /* ind = b8-b15 of lsf[i] */ offset = lsf[i] & 0x00ff; logic16 (); /* offset = b0-b7 of lsf[i] */ /* lsp[i] = table[ind]+ ((table[ind+1]-table[ind])*offset) / 256 */ L_tmp = L_mult (sub (table[ind + 1], table[ind]), offset); lsp[i] = add (table[ind], extract_l (L_shr (L_tmp, 9))); move16 (); } return; } void Lsp_lsf ( Word16 lsp[], /* (i) : lsp[m] (range: -1<=val<1) */ Word16 lsf[], /* (o) : lsf[m] normalized (range: 0.0<=val<=0.5) */ Word16 m /* (i) : LPC order */ ) { Word16 i, ind; Word32 L_tmp; ind = 63; /* begin at end of table -1 */ for (i = m - 1; i >= 0; i--) { /* find value in table that is just greater than lsp[i] */ test (); while (table[ind] < lsp[i]) { ind--; test (); } /* acos(lsp[i])= ind*256 + ( ( lsp[i]-table[ind] ) * slope[ind] )/4096 */ L_tmp = L_mult (sub (lsp[i], table[ind]), slope[ind]); /*(lsp[i]-table[ind])*slope[ind])>>12*/ lsf[i] = round (L_shl (L_tmp, 3)); move16 (); lsf[i] = add (lsf[i], shl (ind, 8)); move16 (); } return; }