view libfaad2/lt_predict.c @ 11623:ecaf7047b6e8

Patch from the author, Zoltan Hidvegi: The filmdint filter does not handle NTSC "telecined" 15fps movies where there is a frame break in the middle of every second NTSC frame, it outputs only 15 frames for every 30 input frames, ignoring the io option. You can notice this during encoding such a sequence you will have lots of diplicate frames / skip frames messages. The patch below fixes this.

/*
** FAAD2 - Freeware Advanced Audio (AAC) Decoder including SBR decoding
** Copyright (C) 2003 M. Bakker, Ahead Software AG, http://www.nero.com
**  
** This program is free software; you can redistribute it and/or modify
** it under the terms of the GNU General Public License as published by
** the Free Software Foundation; either version 2 of the License, or
** (at your option) any later version.
** 
** This program is distributed in the hope that it will be useful,
** but WITHOUT ANY WARRANTY; without even the implied warranty of
** MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
** GNU General Public License for more details.
** 
** You should have received a copy of the GNU General Public License
** along with this program; if not, write to the Free Software 
** Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
**
** Any non-GPL usage of this software or parts of this software is strictly
** forbidden.
**
** Commercial non-GPL licensing of this software is possible.
** For more info contact Ahead Software through [e-mail].
**
** $Id: lt_predict.c,v 1.12 2003/09/09 18:09:52 menno Exp $
**/


#include "common.h"
#include "structs.h"

#ifdef LTP_DEC

#include <stdlib.h>
#include "syntax.h"
#include "lt_predict.h"
#include "filtbank.h"
#include "tns.h"

static real_t codebook[8] =
{
    COEF_CONST(0.570829),
    COEF_CONST(0.696616),
    COEF_CONST(0.813004),
    COEF_CONST(0.911304),
    COEF_CONST(0.984900),
    COEF_CONST(1.067894),
    COEF_CONST(1.194601),
    COEF_CONST(1.369533)
};

void lt_prediction(ic_stream *ics, ltp_info *ltp, real_t *spec,
                   real_t *lt_pred_stat, fb_info *fb, uint8_t win_shape,
                   uint8_t win_shape_prev, uint8_t sr_index,
                   uint8_t object_type, uint16_t frame_len)
{
    uint8_t sfb;
    uint16_t bin, i, num_samples;
    real_t *x_est;
    real_t *X_est;

    if (ics->window_sequence != EIGHT_SHORT_SEQUENCE)
    {
        if (ltp->data_present)
        {
            num_samples = frame_len << 1;

            x_est = (real_t*)malloc(num_samples*sizeof(real_t));
            X_est = (real_t*)malloc(num_samples*sizeof(real_t));

            for(i = 0; i < num_samples; i++)
            {
                /* The extra lookback M (N/2 for LD, 0 for LTP) is handled
                   in the buffer updating */
                x_est[i] = MUL_R_C(lt_pred_stat[num_samples + i - ltp->lag],
                    codebook[ltp->coef]);
            }

            filter_bank_ltp(fb, ics->window_sequence, win_shape, win_shape_prev,
                x_est, X_est, object_type, frame_len);

            tns_encode_frame(ics, &(ics->tns), sr_index, object_type, X_est,
                frame_len);

            for (sfb = 0; sfb < ltp->last_band; sfb++)
            {
                if (ltp->long_used[sfb])
                {
                    uint16_t low  = ics->swb_offset[sfb];
                    uint16_t high = ics->swb_offset[sfb+1];

                    for (bin = low; bin < high; bin++)
                    {
                        spec[bin] += X_est[bin];
                    }
                }
            }

            free(x_est);
            free(X_est);
        }
    }
}

void lt_update_state(real_t *lt_pred_stat, real_t *time, real_t *overlap,
                     uint16_t frame_len, uint8_t object_type)
{
    uint16_t i;

    /*
     * The reference point for index i and the content of the buffer
     * lt_pred_stat are arranged so that lt_pred_stat(0 ... N/2 - 1) contains the
     * last aliased half window from the IMDCT, and lt_pred_stat(N/2 ... N-1)
     * is always all zeros. The rest of lt_pred_stat (i<0) contains the previous
     * fully reconstructed time domain samples, i.e., output of the decoder.
     *
     * These values are shifted up by N*2 to avoid (i<0)
     *
     * For the LD object type an extra 512 samples lookback is accomodated here.
     */
#ifdef LD_DEC
    if (object_type == LD)
    {
        for (i = 0; i < frame_len; i++)
        {
            lt_pred_stat[i]  /* extra 512 */  = lt_pred_stat[i + frame_len];
            lt_pred_stat[frame_len + i]       = lt_pred_stat[i + (frame_len * 2)];
            lt_pred_stat[(frame_len * 2) + i] = time[i];
            lt_pred_stat[(frame_len * 3) + i] = overlap[i];
        }
    } else {
#endif
        for (i = 0; i < frame_len; i++)
        {
            lt_pred_stat[i]                   = lt_pred_stat[i + frame_len];
            lt_pred_stat[frame_len + i]       = time[i];
            lt_pred_stat[(frame_len * 2) + i] = overlap[i];
#if 0 /* set to zero once upon initialisation */
            lt_pred_stat[(frame_len * 3) + i] = 0;
#endif
        }
#ifdef LD_DEC
    }
#endif
}

#endif