FreeCalypso > hg > gsm-codec-lib
view libgsmefr/pre_proc.c @ 242:f081a6850fb5
libgsmfrp: new refined implementation
The previous implementation exhibited the following defects,
which are now fixed:
1) The last received valid SID was cached forever for the purpose of
handling future invalid SIDs - we could have received some valid
SID ages ago, then lots of speech or NO_DATA, and if we then get
an invalid SID, we would resurrect the last valid SID from ancient
history - a bad design. In our new design, we handle invalid SID
based on the current state, much like BFI.
2) GSM 06.11 spec says clearly that after the second lost SID
(received BFI=1 && TAF=1 in CN state) we need to gradually decrease
the output level, rather than jump directly to emitting silence
frames - we previously failed to implement such logic.
3) Per GSM 06.12 section 5.2, Xmaxc should be the same in all 4 subframes
in a SID frame. What should we do if we receive an otherwise valid
SID frame with different Xmaxc? Our previous approach would
replicate this Xmaxc oddity in every subsequent generated CN frame,
which is rather bad. In our new design, the very first CN frame
(which can be seen as a transformation of the SID frame itself)
retains the original 4 distinct Xmaxc, but all subsequent CN frames
are based on the Xmaxc from the last subframe of the most recent SID.
author | Mychaela Falconia <falcon@freecalypso.org> |
---|---|
date | Tue, 09 May 2023 05:16:31 +0000 |
parents | 3ea19a9aa2a1 |
children |
line wrap: on
line source
/************************************************************************* * * FUNCTION: Pre_Process() * * PURPOSE: Preprocessing of input speech. * * DESCRIPTION: * - 2nd order high pass filtering with cut off frequency at 80 Hz. * - Divide input by two. * *************************************************************************/ #include "gsm_efr.h" #include "typedef.h" #include "namespace.h" #include "basic_op.h" #include "oper_32b.h" #include "no_count.h" #include "sig_proc.h" #include "cnst.h" #include "enc_state.h" /*------------------------------------------------------------------------* * * * Algorithm: * * * * y[i] = b[0]*x[i]/2 + b[1]*x[i-1]/2 + b[2]*x[i-2]/2 * * + a[1]*y[i-1] + a[2]*y[i-2]; * * * * * * Input is divided by two in the filtering process. * *------------------------------------------------------------------------*/ /* filter coefficients (fc = 80 Hz, coeff. b[] is divided by 2) */ static const Word16 b[3] = {1899, -3798, 1899}; static const Word16 a[3] = {4096, 7807, -3733}; /* Initialization of static values */ void Init_Pre_Process (struct EFR_encoder_state *st) { struct preproc_state *pps = &st->pre_proc; pps->y2_hi = 0; pps->y2_lo = 0; pps->y1_hi = 0; pps->y1_lo = 0; pps->x0 = 0; pps->x1 = 0; } void Pre_Process ( struct EFR_encoder_state *st, Word16 signal[], /* input/output signal */ Word16 lg) /* lenght of signal */ { struct preproc_state *pps = &st->pre_proc; Word16 i, x2; Word32 L_tmp; for (i = 0; i < lg; i++) { x2 = pps->x1; pps->x1 = pps->x0; pps->x0 = signal[i]; /* y[i] = b[0]*x[i]/2 + b[1]*x[i-1]/2 + b140[2]*x[i-2]/2 */ /* + a[1]*y[i-1] + a[2] * y[i-2]; */ L_tmp = Mpy_32_16 (pps->y1_hi, pps->y1_lo, a[1]); L_tmp = L_add (L_tmp, Mpy_32_16 (pps->y2_hi, pps->y2_lo, a[2])); L_tmp = L_mac (L_tmp, pps->x0, b[0]); L_tmp = L_mac (L_tmp, pps->x1, b[1]); L_tmp = L_mac (L_tmp, x2, b[2]); L_tmp = L_shl (L_tmp, 3); signal[i] = round (L_tmp); pps->y2_hi = pps->y1_hi; pps->y2_lo = pps->y1_lo; L_Extract (L_tmp, &pps->y1_hi, &pps->y1_lo); } return; }