2006-01-03 02:30:38 +00:00
|
|
|
/*
|
|
|
|
* DSP Group TrueSpeech compatible decoder
|
|
|
|
* Copyright (c) 2005 Konstantin Shishkov
|
|
|
|
*
|
2011-03-18 17:35:10 +00:00
|
|
|
* This file is part of Libav.
|
2006-10-07 15:30:46 +00:00
|
|
|
*
|
2011-03-18 17:35:10 +00:00
|
|
|
* Libav is free software; you can redistribute it and/or
|
2006-01-03 02:30:38 +00:00
|
|
|
* modify it under the terms of the GNU Lesser General Public
|
|
|
|
* License as published by the Free Software Foundation; either
|
2006-10-07 15:30:46 +00:00
|
|
|
* version 2.1 of the License, or (at your option) any later version.
|
2006-01-03 02:30:38 +00:00
|
|
|
*
|
2011-03-18 17:35:10 +00:00
|
|
|
* Libav is distributed in the hope that it will be useful,
|
2006-01-03 02:30:38 +00:00
|
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
|
|
* Lesser General Public License for more details.
|
|
|
|
*
|
|
|
|
* You should have received a copy of the GNU Lesser General Public
|
2011-03-18 17:35:10 +00:00
|
|
|
* License along with Libav; if not, write to the Free Software
|
2006-01-12 22:43:26 +00:00
|
|
|
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
2006-01-03 02:30:38 +00:00
|
|
|
*/
|
2009-01-11 22:19:48 +00:00
|
|
|
|
2012-11-10 15:00:00 +00:00
|
|
|
#include "libavutil/channel_layout.h"
|
2009-01-11 22:19:48 +00:00
|
|
|
#include "libavutil/intreadwrite.h"
|
2006-01-03 02:30:38 +00:00
|
|
|
#include "avcodec.h"
|
2011-10-13 04:17:05 +00:00
|
|
|
#include "dsputil.h"
|
|
|
|
#include "get_bits.h"
|
2012-11-10 12:22:56 +00:00
|
|
|
#include "internal.h"
|
2006-01-03 02:30:38 +00:00
|
|
|
|
|
|
|
#include "truespeech_data.h"
|
|
|
|
/**
|
2010-04-20 14:45:34 +00:00
|
|
|
* @file
|
2006-01-03 02:30:38 +00:00
|
|
|
* TrueSpeech decoder.
|
|
|
|
*/
|
|
|
|
|
|
|
|
/**
|
|
|
|
* TrueSpeech decoder context
|
|
|
|
*/
|
|
|
|
typedef struct {
|
2011-10-13 04:17:05 +00:00
|
|
|
DSPContext dsp;
|
2006-01-03 02:30:38 +00:00
|
|
|
/* input data */
|
2012-02-01 05:30:18 +00:00
|
|
|
DECLARE_ALIGNED(16, uint8_t, buffer)[32];
|
2011-08-05 11:15:04 +00:00
|
|
|
int16_t vector[8]; ///< input vector: 5/5/4/4/4/3/3/3
|
|
|
|
int offset1[2]; ///< 8-bit value, used in one copying offset
|
|
|
|
int offset2[4]; ///< 7-bit value, encodes offsets for copying and for two-point filter
|
|
|
|
int pulseoff[4]; ///< 4-bit offset of pulse values block
|
|
|
|
int pulsepos[4]; ///< 27-bit variable, encodes 7 pulse positions
|
|
|
|
int pulseval[4]; ///< 7x2-bit pulse values
|
|
|
|
int flag; ///< 1-bit flag, shows how to choose filters
|
2006-01-03 02:30:38 +00:00
|
|
|
/* temporary data */
|
|
|
|
int filtbuf[146]; // some big vector used for storing filters
|
|
|
|
int prevfilt[8]; // filter from previous frame
|
|
|
|
int16_t tmp1[8]; // coefficients for adding to out
|
|
|
|
int16_t tmp2[8]; // coefficients for adding to out
|
|
|
|
int16_t tmp3[8]; // coefficients for adding to out
|
|
|
|
int16_t cvector[8]; // correlated input vector
|
|
|
|
int filtval; // gain value for one function
|
|
|
|
int16_t newvec[60]; // tmp vector
|
|
|
|
int16_t filters[32]; // filters for every subframe
|
|
|
|
} TSContext;
|
|
|
|
|
2008-03-21 03:11:20 +00:00
|
|
|
static av_cold int truespeech_decode_init(AVCodecContext * avctx)
|
2006-01-03 02:30:38 +00:00
|
|
|
{
|
2011-10-13 04:17:05 +00:00
|
|
|
TSContext *c = avctx->priv_data;
|
2006-01-03 02:30:38 +00:00
|
|
|
|
2011-10-13 03:23:18 +00:00
|
|
|
if (avctx->channels != 1) {
|
2013-03-13 20:17:05 +00:00
|
|
|
avpriv_request_sample(avctx, "Channel count %d", avctx->channels);
|
2012-12-23 17:10:05 +00:00
|
|
|
return AVERROR_PATCHWELCOME;
|
2011-10-13 03:23:18 +00:00
|
|
|
}
|
|
|
|
|
2012-10-23 04:53:16 +00:00
|
|
|
avctx->channel_layout = AV_CH_LAYOUT_MONO;
|
|
|
|
avctx->sample_fmt = AV_SAMPLE_FMT_S16;
|
2011-10-13 04:17:05 +00:00
|
|
|
|
2012-02-15 10:06:44 +00:00
|
|
|
ff_dsputil_init(&c->dsp, avctx);
|
2011-10-13 04:17:05 +00:00
|
|
|
|
2006-01-03 02:30:38 +00:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-02-01 15:57:54 +00:00
|
|
|
static void truespeech_read_frame(TSContext *dec, const uint8_t *input)
|
2006-01-03 02:30:38 +00:00
|
|
|
{
|
2011-10-13 04:17:05 +00:00
|
|
|
GetBitContext gb;
|
|
|
|
|
|
|
|
dec->dsp.bswap_buf((uint32_t *)dec->buffer, (const uint32_t *)input, 8);
|
|
|
|
init_get_bits(&gb, dec->buffer, 32 * 8);
|
|
|
|
|
|
|
|
dec->vector[7] = ts_codebook[7][get_bits(&gb, 3)];
|
|
|
|
dec->vector[6] = ts_codebook[6][get_bits(&gb, 3)];
|
|
|
|
dec->vector[5] = ts_codebook[5][get_bits(&gb, 3)];
|
|
|
|
dec->vector[4] = ts_codebook[4][get_bits(&gb, 4)];
|
|
|
|
dec->vector[3] = ts_codebook[3][get_bits(&gb, 4)];
|
|
|
|
dec->vector[2] = ts_codebook[2][get_bits(&gb, 4)];
|
|
|
|
dec->vector[1] = ts_codebook[1][get_bits(&gb, 5)];
|
|
|
|
dec->vector[0] = ts_codebook[0][get_bits(&gb, 5)];
|
|
|
|
dec->flag = get_bits1(&gb);
|
|
|
|
|
|
|
|
dec->offset1[0] = get_bits(&gb, 4) << 4;
|
|
|
|
dec->offset2[3] = get_bits(&gb, 7);
|
|
|
|
dec->offset2[2] = get_bits(&gb, 7);
|
|
|
|
dec->offset2[1] = get_bits(&gb, 7);
|
|
|
|
dec->offset2[0] = get_bits(&gb, 7);
|
|
|
|
|
|
|
|
dec->offset1[1] = get_bits(&gb, 4);
|
|
|
|
dec->pulseval[1] = get_bits(&gb, 14);
|
|
|
|
dec->pulseval[0] = get_bits(&gb, 14);
|
|
|
|
|
|
|
|
dec->offset1[1] |= get_bits(&gb, 4) << 4;
|
|
|
|
dec->pulseval[3] = get_bits(&gb, 14);
|
|
|
|
dec->pulseval[2] = get_bits(&gb, 14);
|
|
|
|
|
|
|
|
dec->offset1[0] |= get_bits1(&gb);
|
|
|
|
dec->pulsepos[0] = get_bits_long(&gb, 27);
|
|
|
|
dec->pulseoff[0] = get_bits(&gb, 4);
|
|
|
|
|
|
|
|
dec->offset1[0] |= get_bits1(&gb) << 1;
|
|
|
|
dec->pulsepos[1] = get_bits_long(&gb, 27);
|
|
|
|
dec->pulseoff[1] = get_bits(&gb, 4);
|
|
|
|
|
|
|
|
dec->offset1[0] |= get_bits1(&gb) << 2;
|
|
|
|
dec->pulsepos[2] = get_bits_long(&gb, 27);
|
|
|
|
dec->pulseoff[2] = get_bits(&gb, 4);
|
|
|
|
|
|
|
|
dec->offset1[0] |= get_bits1(&gb) << 3;
|
|
|
|
dec->pulsepos[3] = get_bits_long(&gb, 27);
|
|
|
|
dec->pulseoff[3] = get_bits(&gb, 4);
|
2006-01-03 02:30:38 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static void truespeech_correlate_filter(TSContext *dec)
|
|
|
|
{
|
|
|
|
int16_t tmp[8];
|
|
|
|
int i, j;
|
|
|
|
|
|
|
|
for(i = 0; i < 8; i++){
|
|
|
|
if(i > 0){
|
2011-10-13 18:03:17 +00:00
|
|
|
memcpy(tmp, dec->cvector, i * sizeof(*tmp));
|
2006-01-03 02:30:38 +00:00
|
|
|
for(j = 0; j < i; j++)
|
|
|
|
dec->cvector[j] = ((tmp[i - j - 1] * dec->vector[i]) +
|
|
|
|
(dec->cvector[j] << 15) + 0x4000) >> 15;
|
|
|
|
}
|
|
|
|
dec->cvector[i] = (8 - dec->vector[i]) >> 3;
|
|
|
|
}
|
|
|
|
for(i = 0; i < 8; i++)
|
2010-11-06 13:31:17 +00:00
|
|
|
dec->cvector[i] = (dec->cvector[i] * ts_decay_994_1000[i]) >> 15;
|
2006-01-03 02:30:38 +00:00
|
|
|
|
|
|
|
dec->filtval = dec->vector[0];
|
|
|
|
}
|
|
|
|
|
|
|
|
static void truespeech_filters_merge(TSContext *dec)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
|
|
|
if(!dec->flag){
|
|
|
|
for(i = 0; i < 8; i++){
|
|
|
|
dec->filters[i + 0] = dec->prevfilt[i];
|
|
|
|
dec->filters[i + 8] = dec->prevfilt[i];
|
|
|
|
}
|
|
|
|
}else{
|
|
|
|
for(i = 0; i < 8; i++){
|
|
|
|
dec->filters[i + 0]=(dec->cvector[i] * 21846 + dec->prevfilt[i] * 10923 + 16384) >> 15;
|
|
|
|
dec->filters[i + 8]=(dec->cvector[i] * 10923 + dec->prevfilt[i] * 21846 + 16384) >> 15;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
for(i = 0; i < 8; i++){
|
|
|
|
dec->filters[i + 16] = dec->cvector[i];
|
|
|
|
dec->filters[i + 24] = dec->cvector[i];
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void truespeech_apply_twopoint_filter(TSContext *dec, int quart)
|
|
|
|
{
|
2006-06-18 11:33:14 +00:00
|
|
|
int16_t tmp[146 + 60], *ptr0, *ptr1;
|
|
|
|
const int16_t *filter;
|
2006-01-03 02:30:38 +00:00
|
|
|
int i, t, off;
|
|
|
|
|
|
|
|
t = dec->offset2[quart];
|
|
|
|
if(t == 127){
|
2011-10-13 18:03:17 +00:00
|
|
|
memset(dec->newvec, 0, 60 * sizeof(*dec->newvec));
|
2006-01-03 02:30:38 +00:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
for(i = 0; i < 146; i++)
|
|
|
|
tmp[i] = dec->filtbuf[i];
|
|
|
|
off = (t / 25) + dec->offset1[quart >> 1] + 18;
|
2011-12-21 23:13:06 +00:00
|
|
|
off = av_clip(off, 0, 145);
|
2006-01-03 02:30:38 +00:00
|
|
|
ptr0 = tmp + 145 - off;
|
|
|
|
ptr1 = tmp + 146;
|
2012-08-23 18:47:29 +00:00
|
|
|
filter = ts_order2_coeffs + (t % 25) * 2;
|
2006-01-03 02:30:38 +00:00
|
|
|
for(i = 0; i < 60; i++){
|
|
|
|
t = (ptr0[0] * filter[0] + ptr0[1] * filter[1] + 0x2000) >> 14;
|
|
|
|
ptr0++;
|
|
|
|
dec->newvec[i] = t;
|
|
|
|
ptr1[i] = t;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void truespeech_place_pulses(TSContext *dec, int16_t *out, int quart)
|
|
|
|
{
|
|
|
|
int16_t tmp[7];
|
|
|
|
int i, j, t;
|
2006-06-18 11:33:14 +00:00
|
|
|
const int16_t *ptr1;
|
|
|
|
int16_t *ptr2;
|
2006-01-03 02:30:38 +00:00
|
|
|
int coef;
|
|
|
|
|
2011-10-13 18:03:17 +00:00
|
|
|
memset(out, 0, 60 * sizeof(*out));
|
2006-01-03 02:30:38 +00:00
|
|
|
for(i = 0; i < 7; i++) {
|
|
|
|
t = dec->pulseval[quart] & 3;
|
|
|
|
dec->pulseval[quart] >>= 2;
|
2010-11-06 13:31:17 +00:00
|
|
|
tmp[6 - i] = ts_pulse_scales[dec->pulseoff[quart] * 4 + t];
|
2006-01-03 02:30:38 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
coef = dec->pulsepos[quart] >> 15;
|
2012-08-23 18:47:29 +00:00
|
|
|
ptr1 = ts_pulse_values + 30;
|
2006-01-03 02:30:38 +00:00
|
|
|
ptr2 = tmp;
|
|
|
|
for(i = 0, j = 3; (i < 30) && (j > 0); i++){
|
|
|
|
t = *ptr1++;
|
|
|
|
if(coef >= t)
|
|
|
|
coef -= t;
|
|
|
|
else{
|
|
|
|
out[i] = *ptr2++;
|
|
|
|
ptr1 += 30;
|
|
|
|
j--;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
coef = dec->pulsepos[quart] & 0x7FFF;
|
2012-08-23 18:47:29 +00:00
|
|
|
ptr1 = ts_pulse_values;
|
2006-01-03 02:30:38 +00:00
|
|
|
for(i = 30, j = 4; (i < 60) && (j > 0); i++){
|
|
|
|
t = *ptr1++;
|
|
|
|
if(coef >= t)
|
|
|
|
coef -= t;
|
|
|
|
else{
|
|
|
|
out[i] = *ptr2++;
|
|
|
|
ptr1 += 30;
|
|
|
|
j--;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
static void truespeech_update_filters(TSContext *dec, int16_t *out, int quart)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
2011-10-29 22:25:40 +00:00
|
|
|
memmove(dec->filtbuf, &dec->filtbuf[60], 86 * sizeof(*dec->filtbuf));
|
2006-01-03 02:30:38 +00:00
|
|
|
for(i = 0; i < 60; i++){
|
|
|
|
dec->filtbuf[i + 86] = out[i] + dec->newvec[i] - (dec->newvec[i] >> 3);
|
|
|
|
out[i] += dec->newvec[i];
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void truespeech_synth(TSContext *dec, int16_t *out, int quart)
|
|
|
|
{
|
|
|
|
int i,k;
|
|
|
|
int t[8];
|
|
|
|
int16_t *ptr0, *ptr1;
|
|
|
|
|
|
|
|
ptr0 = dec->tmp1;
|
|
|
|
ptr1 = dec->filters + quart * 8;
|
|
|
|
for(i = 0; i < 60; i++){
|
|
|
|
int sum = 0;
|
|
|
|
for(k = 0; k < 8; k++)
|
|
|
|
sum += ptr0[k] * ptr1[k];
|
|
|
|
sum = (sum + (out[i] << 12) + 0x800) >> 12;
|
2007-02-25 10:27:12 +00:00
|
|
|
out[i] = av_clip(sum, -0x7FFE, 0x7FFE);
|
2006-01-03 02:30:38 +00:00
|
|
|
for(k = 7; k > 0; k--)
|
|
|
|
ptr0[k] = ptr0[k - 1];
|
|
|
|
ptr0[0] = out[i];
|
|
|
|
}
|
|
|
|
|
|
|
|
for(i = 0; i < 8; i++)
|
2010-11-06 13:31:17 +00:00
|
|
|
t[i] = (ts_decay_35_64[i] * ptr1[i]) >> 15;
|
2006-01-03 02:30:38 +00:00
|
|
|
|
|
|
|
ptr0 = dec->tmp2;
|
|
|
|
for(i = 0; i < 60; i++){
|
|
|
|
int sum = 0;
|
|
|
|
for(k = 0; k < 8; k++)
|
|
|
|
sum += ptr0[k] * t[k];
|
|
|
|
for(k = 7; k > 0; k--)
|
|
|
|
ptr0[k] = ptr0[k - 1];
|
|
|
|
ptr0[0] = out[i];
|
|
|
|
out[i] = ((out[i] << 12) - sum) >> 12;
|
|
|
|
}
|
|
|
|
|
|
|
|
for(i = 0; i < 8; i++)
|
2010-11-06 13:31:17 +00:00
|
|
|
t[i] = (ts_decay_3_4[i] * ptr1[i]) >> 15;
|
2006-01-03 02:30:38 +00:00
|
|
|
|
|
|
|
ptr0 = dec->tmp3;
|
|
|
|
for(i = 0; i < 60; i++){
|
|
|
|
int sum = out[i] << 12;
|
|
|
|
for(k = 0; k < 8; k++)
|
|
|
|
sum += ptr0[k] * t[k];
|
|
|
|
for(k = 7; k > 0; k--)
|
|
|
|
ptr0[k] = ptr0[k - 1];
|
2007-02-25 10:27:12 +00:00
|
|
|
ptr0[0] = av_clip((sum + 0x800) >> 12, -0x7FFE, 0x7FFE);
|
2006-01-03 02:30:38 +00:00
|
|
|
|
|
|
|
sum = ((ptr0[1] * (dec->filtval - (dec->filtval >> 2))) >> 4) + sum;
|
|
|
|
sum = sum - (sum >> 3);
|
2007-02-25 10:27:12 +00:00
|
|
|
out[i] = av_clip((sum + 0x800) >> 12, -0x7FFE, 0x7FFE);
|
2006-01-03 02:30:38 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void truespeech_save_prevvec(TSContext *c)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
|
|
|
for(i = 0; i < 8; i++)
|
|
|
|
c->prevfilt[i] = c->cvector[i];
|
|
|
|
}
|
|
|
|
|
2011-09-06 16:17:45 +00:00
|
|
|
static int truespeech_decode_frame(AVCodecContext *avctx, void *data,
|
|
|
|
int *got_frame_ptr, AVPacket *avpkt)
|
2006-01-03 02:30:38 +00:00
|
|
|
{
|
2012-12-24 01:04:31 +00:00
|
|
|
AVFrame *frame = data;
|
2009-04-07 15:59:50 +00:00
|
|
|
const uint8_t *buf = avpkt->data;
|
|
|
|
int buf_size = avpkt->size;
|
2006-01-03 02:30:38 +00:00
|
|
|
TSContext *c = avctx->priv_data;
|
|
|
|
|
2007-09-13 05:59:58 +00:00
|
|
|
int i, j;
|
2011-09-06 16:17:45 +00:00
|
|
|
int16_t *samples;
|
|
|
|
int iterations, ret;
|
2006-01-03 02:30:38 +00:00
|
|
|
|
2011-10-13 03:15:00 +00:00
|
|
|
iterations = buf_size / 32;
|
|
|
|
|
|
|
|
if (!iterations) {
|
2010-07-11 14:50:20 +00:00
|
|
|
av_log(avctx, AV_LOG_ERROR,
|
|
|
|
"Too small input buffer (%d bytes), need at least 32 bytes\n", buf_size);
|
|
|
|
return -1;
|
|
|
|
}
|
2011-10-13 03:15:00 +00:00
|
|
|
|
2011-09-06 16:17:45 +00:00
|
|
|
/* get output buffer */
|
2012-12-24 01:04:31 +00:00
|
|
|
frame->nb_samples = iterations * 240;
|
2012-11-21 20:34:46 +00:00
|
|
|
if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) {
|
2011-09-06 16:17:45 +00:00
|
|
|
av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
|
|
|
|
return ret;
|
2011-10-13 03:15:00 +00:00
|
|
|
}
|
2012-12-24 01:04:31 +00:00
|
|
|
samples = (int16_t *)frame->data[0];
|
2011-10-13 03:15:00 +00:00
|
|
|
|
2011-09-06 16:17:45 +00:00
|
|
|
memset(samples, 0, iterations * 240 * sizeof(*samples));
|
2011-10-13 03:27:32 +00:00
|
|
|
|
2007-09-13 05:59:58 +00:00
|
|
|
for(j = 0; j < iterations; j++) {
|
2011-10-13 04:21:17 +00:00
|
|
|
truespeech_read_frame(c, buf);
|
|
|
|
buf += 32;
|
2006-01-03 02:30:38 +00:00
|
|
|
|
|
|
|
truespeech_correlate_filter(c);
|
|
|
|
truespeech_filters_merge(c);
|
|
|
|
|
|
|
|
for(i = 0; i < 4; i++) {
|
|
|
|
truespeech_apply_twopoint_filter(c, i);
|
2011-10-13 03:27:32 +00:00
|
|
|
truespeech_place_pulses (c, samples, i);
|
|
|
|
truespeech_update_filters(c, samples, i);
|
|
|
|
truespeech_synth (c, samples, i);
|
|
|
|
samples += 60;
|
2006-01-03 02:30:38 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
truespeech_save_prevvec(c);
|
|
|
|
}
|
|
|
|
|
2012-12-24 01:04:31 +00:00
|
|
|
*got_frame_ptr = 1;
|
2006-01-03 02:30:38 +00:00
|
|
|
|
2011-10-13 04:21:17 +00:00
|
|
|
return buf_size;
|
2006-01-03 02:30:38 +00:00
|
|
|
}
|
|
|
|
|
2011-01-25 21:40:11 +00:00
|
|
|
AVCodec ff_truespeech_decoder = {
|
2011-07-17 10:54:31 +00:00
|
|
|
.name = "truespeech",
|
2013-10-03 20:57:53 +00:00
|
|
|
.long_name = NULL_IF_CONFIG_SMALL("DSP Group TrueSpeech"),
|
2011-07-17 10:54:31 +00:00
|
|
|
.type = AVMEDIA_TYPE_AUDIO,
|
2012-08-05 09:11:04 +00:00
|
|
|
.id = AV_CODEC_ID_TRUESPEECH,
|
2011-07-17 10:54:31 +00:00
|
|
|
.priv_data_size = sizeof(TSContext),
|
|
|
|
.init = truespeech_decode_init,
|
|
|
|
.decode = truespeech_decode_frame,
|
2011-09-06 16:17:45 +00:00
|
|
|
.capabilities = CODEC_CAP_DR1,
|
2006-01-03 02:30:38 +00:00
|
|
|
};
|