* Copyright (c) 2008 Vladimir Voroshilov
* Copyright (c) 2009 Vitor Sessak
*
- * This file is part of FFmpeg.
+ * This file is part of Libav.
*
- * FFmpeg is free software; you can redistribute it and/or
+ * Libav is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
- * FFmpeg is distributed in the hope that it will be useful,
+ * Libav 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
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
- * License along with FFmpeg; if not, write to the Free Software
+ * License along with Libav; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include <math.h>
#include <stdint.h>
+#include <string.h>
-#include "avcodec.h"
-#define ALT_BITSTREAM_READER_LE
-#include "get_bits.h"
-#include "dsputil.h"
+#include "libavutil/channel_layout.h"
+#include "libavutil/float_dsp.h"
+#include "libavutil/mathematics.h"
+#define BITSTREAM_READER_LE
+#include "avcodec.h"
+#include "bitstream.h"
+#include "internal.h"
#include "lsp.h"
-#include "celp_math.h"
#include "acelp_vectors.h"
#include "acelp_pitch_delay.h"
#include "acelp_filters.h"
#include "celp_filters.h"
-#define LSFQ_DIFF_MIN (0.0125 * M_PI)
-
-#define LP_FILTER_ORDER 10
-
-/** Number of past samples needed for excitation interpolation */
-#define L_INTERPOL (LP_FILTER_ORDER + 1)
-
-/** Subframe size for all modes except 16k */
-#define SUBFR_SIZE 48
-
#define MAX_SUBFRAME_COUNT 5
+#include "sipr.h"
#include "siprdata.h"
-typedef enum {
- MODE_16k,
- MODE_8k5,
- MODE_6k5,
- MODE_5k0,
- MODE_COUNT
-} SiprMode;
-
-typedef struct {
+typedef struct SiprModeParam {
const char *mode_name;
uint16_t bits_per_frame;
uint8_t subframe_count;
/* bitstream parameters */
uint8_t number_of_fc_indexes;
+ uint8_t ma_predictor_bits; ///< size in bits of the switched MA predictor
/** size in bits of the i-th stage vector of quantizer */
uint8_t vq_indexes_bits[5];
} SiprModeParam;
static const SiprModeParam modes[MODE_COUNT] = {
+ [MODE_16k] = {
+ .mode_name = "16k",
+ .bits_per_frame = 160,
+ .subframe_count = SUBFRAME_COUNT_16k,
+ .frames_per_packet = 1,
+ .pitch_sharp_factor = 0.00,
+
+ .number_of_fc_indexes = 10,
+ .ma_predictor_bits = 1,
+ .vq_indexes_bits = {7, 8, 7, 7, 7},
+ .pitch_delay_bits = {9, 6},
+ .gp_index_bits = 4,
+ .fc_index_bits = {4, 5, 4, 5, 4, 5, 4, 5, 4, 5},
+ .gc_index_bits = 5
+ },
+
[MODE_8k5] = {
.mode_name = "8k5",
.bits_per_frame = 152,
.pitch_sharp_factor = 0.8,
.number_of_fc_indexes = 3,
+ .ma_predictor_bits = 0,
.vq_indexes_bits = {6, 7, 7, 7, 5},
.pitch_delay_bits = {8, 5, 5},
.gp_index_bits = 0,
.pitch_sharp_factor = 0.8,
.number_of_fc_indexes = 3,
+ .ma_predictor_bits = 0,
.vq_indexes_bits = {6, 7, 7, 7, 5},
.pitch_delay_bits = {8, 5, 5},
.gp_index_bits = 0,
.pitch_sharp_factor = 0.85,
.number_of_fc_indexes = 1,
+ .ma_predictor_bits = 0,
.vq_indexes_bits = {6, 7, 7, 7, 5},
.pitch_delay_bits = {8, 5, 8, 5, 5},
.gp_index_bits = 0,
}
};
-typedef struct {
- AVCodecContext *avctx;
- DSPContext dsp;
-
- SiprModeParam m;
- SiprMode mode;
-
- float past_pitch_gain;
- float lsf_history[LP_FILTER_ORDER];
-
- float excitation[L_INTERPOL + PITCH_DELAY_MAX + 5*SUBFR_SIZE];
-
- DECLARE_ALIGNED_16(float, synth_buf[LP_FILTER_ORDER + 5*SUBFR_SIZE + 6]);
-
- float lsp_history[LP_FILTER_ORDER];
- float gain_mem;
- float energy_history[4];
- float highpass_filt_mem[2];
- float postfilter_mem[PITCH_DELAY_MAX + LP_FILTER_ORDER];
-
- /* 5k0 */
- float tilt_mem;
- float postfilter_agc;
- float postfilter_mem5k0[PITCH_DELAY_MAX + LP_FILTER_ORDER];
- float postfilter_syn5k0[LP_FILTER_ORDER + SUBFR_SIZE*5];
-} SiprContext;
-
-typedef struct {
- int vq_indexes[5];
- int pitch_delay[5]; ///< pitch delay
- int gp_index[5]; ///< adaptive-codebook gain indexes
- int16_t fc_indexes[5][10]; ///< fixed-codebook indexes
- int gc_index[5]; ///< fixed-codebook gain indexes
-} SiprParameters;
-
+const float ff_pow_0_5[] = {
+ 1.0/(1 << 1), 1.0/(1 << 2), 1.0/(1 << 3), 1.0/(1 << 4),
+ 1.0/(1 << 5), 1.0/(1 << 6), 1.0/(1 << 7), 1.0/(1 << 8),
+ 1.0/(1 << 9), 1.0/(1 << 10), 1.0/(1 << 11), 1.0/(1 << 12),
+ 1.0/(1 << 13), 1.0/(1 << 14), 1.0/(1 << 15), 1.0/(1 << 16)
+};
-static void dequant(float *out, const int *idx, const float *cbs[])
+static void dequant(float *out, const int *idx, const float * const cbs[])
{
int i;
int stride = 2;
}
/**
- * Extracts decoding parameters from the input bitstream.
+ * Extract decoding parameters from the input bitstream.
* @param parms parameters structure
- * @param pgb pointer to initialized GetBitContext structure
+ * @param bc pointer to initialized BitstreamContext structure
*/
-static void decode_parameters(SiprParameters* parms, GetBitContext *pgb,
+static void decode_parameters(SiprParameters* parms, BitstreamContext *bc,
const SiprModeParam *p)
{
int i, j;
+ if (p->ma_predictor_bits)
+ parms->ma_pred_switch = bitstream_read(bc, p->ma_predictor_bits);
+
for (i = 0; i < 5; i++)
- parms->vq_indexes[i] = get_bits(pgb, p->vq_indexes_bits[i]);
+ parms->vq_indexes[i] = bitstream_read(bc, p->vq_indexes_bits[i]);
for (i = 0; i < p->subframe_count; i++) {
- parms->pitch_delay[i] = get_bits(pgb, p->pitch_delay_bits[i]);
- parms->gp_index[i] = get_bits(pgb, p->gp_index_bits);
+ parms->pitch_delay[i] = bitstream_read(bc, p->pitch_delay_bits[i]);
+ if (p->gp_index_bits)
+ parms->gp_index[i] = bitstream_read(bc, p->gp_index_bits);
for (j = 0; j < p->number_of_fc_indexes; j++)
- parms->fc_indexes[i][j] = get_bits(pgb, p->fc_index_bits[j]);
+ parms->fc_indexes[i][j] = bitstream_read(bc, p->fc_index_bits[j]);
- parms->gc_index[i] = get_bits(pgb, p->gc_index_bits);
+ parms->gc_index[i] = bitstream_read(bc, p->gc_index_bits);
}
}
-static void lsp2lpc_sipr(const double *lsp, float *Az)
-{
- int lp_half_order = LP_FILTER_ORDER >> 1;
- double buf[(LP_FILTER_ORDER >> 1) + 1];
- double pa[(LP_FILTER_ORDER >> 1) + 1];
- double *qa = buf + 1;
- int i,j;
-
- qa[-1] = 0.0;
-
- ff_lsp2polyf(lsp , pa, lp_half_order );
- ff_lsp2polyf(lsp + 1, qa, lp_half_order - 1);
-
- for (i = 1, j = LP_FILTER_ORDER - 1; i < lp_half_order; i++, j--) {
- double paf = pa[i] * (1 + lsp[LP_FILTER_ORDER - 1]);
- double qaf = (qa[i] - qa[i-2]) * (1 - lsp[LP_FILTER_ORDER - 1]);
- Az[i-1] = (paf + qaf) * 0.5;
- Az[j-1] = (paf - qaf) * 0.5;
- }
-
- Az[lp_half_order - 1] = (1.0 + lsp[LP_FILTER_ORDER - 1]) *
- pa[lp_half_order] * 0.5;
-
- Az[LP_FILTER_ORDER - 1] = lsp[LP_FILTER_ORDER - 1];
-}
-
static void sipr_decode_lp(float *lsfnew, const float *lsfold, float *Az,
int num_subfr)
{
for (j = 0; j < LP_FILTER_ORDER; j++)
lsfint[j] = lsfold[j] * (1 - t) + t * lsfnew[j];
- lsp2lpc_sipr(lsfint, Az);
+ ff_amrwb_lsp2lpc(lsfint, Az, LP_FILTER_ORDER);
Az += LP_FILTER_ORDER;
t += t0;
}
}
/**
- * Evaluates the adaptative impulse response.
+ * Evaluate the adaptive impulse response.
*/
static void eval_ir(const float *Az, int pitch_lag, float *freq,
float pitch_sharp_factor)
float tmp1[SUBFR_SIZE+1], tmp2[LP_FILTER_ORDER+1];
int i;
- tmp1[0] = 1.;
+ tmp1[0] = 1.0;
for (i = 0; i < LP_FILTER_ORDER; i++) {
tmp1[i+1] = Az[i] * ff_pow_0_55[i];
tmp2[i ] = Az[i] * ff_pow_0_7 [i];
}
/**
- * Evaluates the convolution of a vector with a sparse vector.
+ * Evaluate the convolution of a vector with a sparse vector.
*/
static void convolute_with_sparse(float *out, const AMRFixed *pulses,
const float *shape, int length)
for (i = 0; i < LP_FILTER_ORDER; i++) {
lpc_d[i] = lpc[i] * ff_pow_0_75[i];
- lpc_n[i] = lpc[i] * pow_0_5 [i];
+ lpc_n[i] = lpc[i] * ff_pow_0_5 [i];
};
memcpy(pole_out - LP_FILTER_ORDER, ctx->postfilter_mem,
float *out_data)
{
int i, j;
- int frame_size = ctx->m.subframe_count * SUBFR_SIZE;
+ int subframe_count = modes[ctx->mode].subframe_count;
+ int frame_size = subframe_count * SUBFR_SIZE;
float Az[LP_FILTER_ORDER * MAX_SUBFRAME_COUNT];
float *excitation;
float ir_buf[SUBFR_SIZE + LP_FILTER_ORDER];
memset(ir_buf, 0, LP_FILTER_ORDER * sizeof(float));
lsf_decode_fp(lsf_new, ctx->lsf_history, params);
- sipr_decode_lp(lsf_new, ctx->lsp_history, Az, ctx->m.subframe_count);
+ sipr_decode_lp(lsf_new, ctx->lsp_history, Az, subframe_count);
memcpy(ctx->lsp_history, lsf_new, LP_FILTER_ORDER * sizeof(float));
excitation = ctx->excitation + PITCH_DELAY_MAX + L_INTERPOL;
- for (i = 0; i < ctx->m.subframe_count; i++) {
+ for (i = 0; i < subframe_count; i++) {
float *pAz = Az + i*LP_FILTER_ORDER;
float fixed_vector[SUBFR_SIZE];
int T0,T0_frac;
decode_fixed_sparse(&fixed_cb, params->fc_indexes[i], ctx->mode,
ctx->past_pitch_gain < 0.8);
- eval_ir(pAz, T0, impulse_response, ctx->m.pitch_sharp_factor);
+ eval_ir(pAz, T0, impulse_response, modes[ctx->mode].pitch_sharp_factor);
convolute_with_sparse(fixed_vector, &fixed_cb, impulse_response,
SUBFR_SIZE);
- avg_energy =
- (0.01 + ff_dot_productf(fixed_vector, fixed_vector, SUBFR_SIZE))/
- SUBFR_SIZE;
+ avg_energy = (0.01 + avpriv_scalarproduct_float_c(fixed_vector,
+ fixed_vector,
+ SUBFR_SIZE)) /
+ SUBFR_SIZE;
ctx->past_pitch_gain = pitch_gain = gain_cb[params->gc_index[i]][0];
gain_code = ff_amr_set_fixed_gain(gain_cb[params->gc_index[i]][1],
avg_energy, ctx->energy_history,
- 34 - 15.0/(log2f(10.0) * 0.05),
+ 34 - 15.0/(0.05*M_LN10/M_LN2),
pred);
ff_weighted_vector_sumf(excitation, excitation, fixed_vector,
LP_FILTER_ORDER * sizeof(float));
if (ctx->mode == MODE_5k0) {
- for (i = 0; i < ctx->m.subframe_count; i++) {
- float energy = ff_dot_productf(ctx->postfilter_syn5k0 + LP_FILTER_ORDER + i*SUBFR_SIZE,
- ctx->postfilter_syn5k0 + LP_FILTER_ORDER + i*SUBFR_SIZE,
- SUBFR_SIZE);
- ff_adaptative_gain_control(&synth[i * SUBFR_SIZE], energy,
- SUBFR_SIZE, 0.9, &ctx->postfilter_agc);
+ for (i = 0; i < subframe_count; i++) {
+ float energy = avpriv_scalarproduct_float_c(ctx->postfilter_syn5k0 + LP_FILTER_ORDER + i * SUBFR_SIZE,
+ ctx->postfilter_syn5k0 + LP_FILTER_ORDER + i * SUBFR_SIZE,
+ SUBFR_SIZE);
+ ff_adaptive_gain_control(&synth[i * SUBFR_SIZE],
+ &synth[i * SUBFR_SIZE], energy,
+ SUBFR_SIZE, 0.9, &ctx->postfilter_agc);
}
memcpy(ctx->postfilter_syn5k0, ctx->postfilter_syn5k0 + frame_size,
LP_FILTER_ORDER*sizeof(float));
}
- memcpy(ctx->excitation, excitation - PITCH_DELAY_MAX - L_INTERPOL,
+ memmove(ctx->excitation, excitation - PITCH_DELAY_MAX - L_INTERPOL,
(PITCH_DELAY_MAX + L_INTERPOL) * sizeof(float));
- ff_acelp_apply_order_2_transfer_function(synth,
+ ff_acelp_apply_order_2_transfer_function(out_data, synth,
(const float[2]) {-1.99997 , 1.000000000},
(const float[2]) {-1.93307352, 0.935891986},
0.939805806,
ctx->highpass_filt_mem,
frame_size);
-
- ctx->dsp.vector_clipf(out_data, synth, -1, 32767./(1<<15), frame_size);
-
}
static av_cold int sipr_decoder_init(AVCodecContext * avctx)
SiprContext *ctx = avctx->priv_data;
int i;
- if (avctx->bit_rate > 12200) ctx->mode = MODE_16k;
- else if (avctx->bit_rate > 7500 ) ctx->mode = MODE_8k5;
- else if (avctx->bit_rate > 5750 ) ctx->mode = MODE_6k5;
- else ctx->mode = MODE_5k0;
+ switch (avctx->block_align) {
+ case 20: ctx->mode = MODE_16k; break;
+ case 19: ctx->mode = MODE_8k5; break;
+ case 29: ctx->mode = MODE_6k5; break;
+ case 37: ctx->mode = MODE_5k0; break;
+ default:
+ if (avctx->bit_rate > 12200) ctx->mode = MODE_16k;
+ else if (avctx->bit_rate > 7500 ) ctx->mode = MODE_8k5;
+ else if (avctx->bit_rate > 5750 ) ctx->mode = MODE_6k5;
+ else ctx->mode = MODE_5k0;
+ av_log(avctx, AV_LOG_WARNING,
+ "Invalid block_align: %d. Mode %s guessed based on bitrate: %d\n",
+ avctx->block_align, modes[ctx->mode].mode_name, avctx->bit_rate);
+ }
+
+ av_log(avctx, AV_LOG_DEBUG, "Mode: %s\n", modes[ctx->mode].mode_name);
- ctx->m = modes[ctx->mode];
- av_log(avctx, AV_LOG_DEBUG, "Mode: %s\n", ctx->m.mode_name);
+ if (ctx->mode == MODE_16k) {
+ ff_sipr_init_16k(ctx);
+ ctx->decode_frame = ff_sipr_decode_frame_16k;
+ } else {
+ ctx->decode_frame = decode_frame;
+ }
for (i = 0; i < LP_FILTER_ORDER; i++)
ctx->lsp_history[i] = cos((i+1) * M_PI / (LP_FILTER_ORDER + 1));
for (i = 0; i < 4; i++)
ctx->energy_history[i] = -14;
- avctx->sample_fmt = SAMPLE_FMT_FLT;
-
- if (ctx->mode == MODE_16k) {
- av_log(avctx, AV_LOG_ERROR, "decoding 16kbps SIPR files is not "
- "supported yet.\n");
- return -1;
- }
-
- dsputil_init(&ctx->dsp, avctx);
+ avctx->channels = 1;
+ avctx->channel_layout = AV_CH_LAYOUT_MONO;
+ avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
return 0;
}
-static int sipr_decode_frame(AVCodecContext *avctx, void *datap,
- int *data_size, AVPacket *avpkt)
+static int sipr_decode_frame(AVCodecContext *avctx, void *data,
+ int *got_frame_ptr, AVPacket *avpkt)
{
SiprContext *ctx = avctx->priv_data;
+ AVFrame *frame = data;
const uint8_t *buf=avpkt->data;
SiprParameters parm;
- GetBitContext gb;
- float *data = datap;
- int i;
+ const SiprModeParam *mode_par = &modes[ctx->mode];
+ BitstreamContext bc;
+ float *samples;
+ int subframe_size = ctx->mode == MODE_16k ? L_SUBFR_16k : SUBFR_SIZE;
+ int i, ret;
ctx->avctx = avctx;
- if (avpkt->size < (ctx->m.bits_per_frame >> 3)) {
+ if (avpkt->size < (mode_par->bits_per_frame >> 3)) {
av_log(avctx, AV_LOG_ERROR,
"Error processing packet: packet size (%d) too small\n",
avpkt->size);
-
- *data_size = 0;
return -1;
}
- if (*data_size < SUBFR_SIZE * ctx->m.subframe_count * sizeof(float)) {
- av_log(avctx, AV_LOG_ERROR,
- "Error processing packet: output buffer (%d) too small\n",
- *data_size);
- *data_size = 0;
- return -1;
+ /* get output buffer */
+ frame->nb_samples = mode_par->frames_per_packet * subframe_size *
+ mode_par->subframe_count;
+ if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) {
+ av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
+ return ret;
}
+ samples = (float *)frame->data[0];
- init_get_bits(&gb, buf, ctx->m.bits_per_frame);
+ bitstream_init(&bc, buf, mode_par->bits_per_frame);
- for (i = 0; i < ctx->m.frames_per_packet; i++) {
- decode_parameters(&parm, &gb, &ctx->m);
- decode_frame(ctx, &parm, data);
+ for (i = 0; i < mode_par->frames_per_packet; i++) {
+ decode_parameters(&parm, &bc, mode_par);
- data += SUBFR_SIZE * ctx->m.subframe_count;
+ ctx->decode_frame(ctx, &parm, samples);
+
+ samples += subframe_size * mode_par->subframe_count;
}
- *data_size = ctx->m.frames_per_packet * SUBFR_SIZE *
- ctx->m.subframe_count * sizeof(float);
+ *got_frame_ptr = 1;
- return ctx->m.bits_per_frame >> 3;
-};
+ return mode_par->bits_per_frame >> 3;
+}
-AVCodec sipr_decoder = {
- "sipr",
- CODEC_TYPE_AUDIO,
- CODEC_ID_SIPR,
- sizeof(SiprContext),
- sipr_decoder_init,
- NULL,
- NULL,
- sipr_decode_frame,
- .long_name = NULL_IF_CONFIG_SMALL("RealAudio SIPR / ACELP.NET"),
+AVCodec ff_sipr_decoder = {
+ .name = "sipr",
+ .long_name = NULL_IF_CONFIG_SMALL("RealAudio SIPR / ACELP.NET"),
+ .type = AVMEDIA_TYPE_AUDIO,
+ .id = AV_CODEC_ID_SIPR,
+ .priv_data_size = sizeof(SiprContext),
+ .init = sipr_decoder_init,
+ .decode = sipr_decode_frame,
+ .capabilities = AV_CODEC_CAP_DR1,
};