sipr.c 18.4 KB
Newer Older
1 2 3 4 5 6
/*
 * SIPR / ACELP.NET decoder
 *
 * Copyright (c) 2008 Vladimir Voroshilov
 * Copyright (c) 2009 Vitor Sessak
 *
7
 * This file is part of Libav.
8
 *
9
 * Libav is free software; you can redistribute it and/or
10 11 12 13
 * 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.
 *
14
 * Libav is distributed in the hope that it will be useful,
15 16 17 18 19
 * 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
20
 * License along with Libav; if not, write to the Free Software
21 22 23 24 25
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

#include <math.h>
#include <stdint.h>
26
#include <string.h>
27

28
#include "libavutil/mathematics.h"
29
#include "avcodec.h"
30
#define BITSTREAM_READER_LE
31 32 33 34 35 36 37 38 39 40
#include "get_bits.h"
#include "dsputil.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"

41 42
#define MAX_SUBFRAME_COUNT   5

43
#include "sipr.h"
44 45 46 47 48 49 50 51 52 53 54
#include "siprdata.h"

typedef struct {
    const char *mode_name;
    uint16_t bits_per_frame;
    uint8_t subframe_count;
    uint8_t frames_per_packet;
    float pitch_sharp_factor;

    /* bitstream parameters */
    uint8_t number_of_fc_indexes;
Vitor Sessak's avatar
Vitor Sessak committed
55
    uint8_t ma_predictor_bits;  ///< size in bits of the switched MA predictor
56 57 58 59 60 61 62 63 64 65 66 67 68

    /** size in bits of the i-th stage vector of quantizer */
    uint8_t vq_indexes_bits[5];

    /** size in bits of the adaptive-codebook index for every subframe */
    uint8_t pitch_delay_bits[5];

    uint8_t gp_index_bits;
    uint8_t fc_index_bits[10]; ///< size in bits of the fixed codebook indexes
    uint8_t gc_index_bits;     ///< size in bits of the gain  codebook indexes
} SiprModeParam;

static const SiprModeParam modes[MODE_COUNT] = {
Vitor Sessak's avatar
Vitor Sessak committed
69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84
    [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
    },

85 86 87 88 89 90 91 92
    [MODE_8k5] = {
        .mode_name          = "8k5",
        .bits_per_frame     = 152,
        .subframe_count     = 3,
        .frames_per_packet  = 1,
        .pitch_sharp_factor = 0.8,

        .number_of_fc_indexes = 3,
Vitor Sessak's avatar
Vitor Sessak committed
93
        .ma_predictor_bits    = 0,
94 95 96 97 98 99 100 101 102 103 104 105 106 107 108
        .vq_indexes_bits      = {6, 7, 7, 7, 5},
        .pitch_delay_bits     = {8, 5, 5},
        .gp_index_bits        = 0,
        .fc_index_bits        = {9, 9, 9},
        .gc_index_bits        = 7
    },

    [MODE_6k5] = {
        .mode_name          = "6k5",
        .bits_per_frame     = 232,
        .subframe_count     = 3,
        .frames_per_packet  = 2,
        .pitch_sharp_factor = 0.8,

        .number_of_fc_indexes = 3,
Vitor Sessak's avatar
Vitor Sessak committed
109
        .ma_predictor_bits    = 0,
110 111 112 113 114 115 116 117 118 119 120 121 122 123 124
        .vq_indexes_bits      = {6, 7, 7, 7, 5},
        .pitch_delay_bits     = {8, 5, 5},
        .gp_index_bits        = 0,
        .fc_index_bits        = {5, 5, 5},
        .gc_index_bits        = 7
    },

    [MODE_5k0] = {
        .mode_name          = "5k0",
        .bits_per_frame     = 296,
        .subframe_count     = 5,
        .frames_per_packet  = 2,
        .pitch_sharp_factor = 0.85,

        .number_of_fc_indexes = 1,
Vitor Sessak's avatar
Vitor Sessak committed
125
        .ma_predictor_bits    = 0,
126 127 128 129 130 131 132 133
        .vq_indexes_bits      = {6, 7, 7, 7, 5},
        .pitch_delay_bits     = {8, 5, 8, 5, 5},
        .gp_index_bits        = 0,
        .fc_index_bits        = {10},
        .gc_index_bits        = 7
    }
};

134 135 136 137 138 139 140
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)
};

141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187
static void dequant(float *out, const int *idx, const float *cbs[])
{
    int i;
    int stride  = 2;
    int num_vec = 5;

    for (i = 0; i < num_vec; i++)
        memcpy(out + stride*i, cbs[i] + stride*idx[i], stride*sizeof(float));

}

static void lsf_decode_fp(float *lsfnew, float *lsf_history,
                          const SiprParameters *parm)
{
    int i;
    float lsf_tmp[LP_FILTER_ORDER];

    dequant(lsf_tmp, parm->vq_indexes, lsf_codebooks);

    for (i = 0; i < LP_FILTER_ORDER; i++)
        lsfnew[i] = lsf_history[i] * 0.33 + lsf_tmp[i] + mean_lsf[i];

    ff_sort_nearly_sorted_floats(lsfnew, LP_FILTER_ORDER - 1);

    /* Note that a minimum distance is not enforced between the last value and
       the previous one, contrary to what is done in ff_acelp_reorder_lsf() */
    ff_set_min_dist_lsf(lsfnew, LSFQ_DIFF_MIN, LP_FILTER_ORDER - 1);
    lsfnew[9] = FFMIN(lsfnew[LP_FILTER_ORDER - 1], 1.3 * M_PI);

    memcpy(lsf_history, lsf_tmp, LP_FILTER_ORDER * sizeof(*lsf_history));

    for (i = 0; i < LP_FILTER_ORDER - 1; i++)
        lsfnew[i] = cos(lsfnew[i]);
    lsfnew[LP_FILTER_ORDER - 1] *= 6.153848 / M_PI;
}

/** Apply pitch lag to the fixed vector (AMR section 6.1.2). */
static void pitch_sharpening(int pitch_lag_int, float beta,
                             float *fixed_vector)
{
    int i;

    for (i = pitch_lag_int; i < SUBFR_SIZE; i++)
        fixed_vector[i] += beta * fixed_vector[i - pitch_lag_int];
}

/**
188
 * Extract decoding parameters from the input bitstream.
189 190 191 192 193 194 195 196
 * @param parms          parameters structure
 * @param pgb            pointer to initialized GetBitContext structure
 */
static void decode_parameters(SiprParameters* parms, GetBitContext *pgb,
                              const SiprModeParam *p)
{
    int i, j;

197 198
    if (p->ma_predictor_bits)
        parms->ma_pred_switch       = get_bits(pgb, p->ma_predictor_bits);
Vitor Sessak's avatar
Vitor Sessak committed
199

200 201 202 203 204
    for (i = 0; i < 5; i++)
        parms->vq_indexes[i]        = get_bits(pgb, 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]);
205 206
        if (p->gp_index_bits)
            parms->gp_index[i]      = get_bits(pgb, p->gp_index_bits);
207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226

        for (j = 0; j < p->number_of_fc_indexes; j++)
            parms->fc_indexes[i][j] = get_bits(pgb, p->fc_index_bits[j]);

        parms->gc_index[i]          = get_bits(pgb, p->gc_index_bits);
    }
}

static void sipr_decode_lp(float *lsfnew, const float *lsfold, float *Az,
                           int num_subfr)
{
    double lsfint[LP_FILTER_ORDER];
    int i,j;
    float t, t0 = 1.0 / num_subfr;

    t = t0 * 0.5;
    for (i = 0; i < num_subfr; i++) {
        for (j = 0; j < LP_FILTER_ORDER; j++)
            lsfint[j] = lsfold[j] * (1 - t) + t * lsfnew[j];

227
        ff_amrwb_lsp2lpc(lsfint, Az, LP_FILTER_ORDER);
228 229 230 231 232 233
        Az += LP_FILTER_ORDER;
        t += t0;
    }
}

/**
234
 * Evaluate the adaptive impulse response.
235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255
 */
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.;
    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];
    }
    memset(tmp1 + 11, 0, 37 * sizeof(float));

    ff_celp_lp_synthesis_filterf(freq, tmp2, tmp1, SUBFR_SIZE,
                                 LP_FILTER_ORDER);

    pitch_sharpening(pitch_lag, pitch_sharp_factor, freq);
}

/**
256
 * Evaluate the convolution of a vector with a sparse vector.
257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281
 */
static void convolute_with_sparse(float *out, const AMRFixed *pulses,
                                  const float *shape, int length)
{
    int i, j;

    memset(out, 0, length*sizeof(float));
    for (i = 0; i < pulses->n; i++)
        for (j = pulses->x[i]; j < length; j++)
            out[j] += pulses->y[i] * shape[j - pulses->x[i]];
}

/**
 * Apply postfilter, very similar to AMR one.
 */
static void postfilter_5k0(SiprContext *ctx, const float *lpc, float *samples)
{
    float buf[SUBFR_SIZE + LP_FILTER_ORDER];
    float *pole_out = buf + LP_FILTER_ORDER;
    float lpc_n[LP_FILTER_ORDER];
    float lpc_d[LP_FILTER_ORDER];
    int i;

    for (i = 0; i < LP_FILTER_ORDER; i++) {
        lpc_d[i] = lpc[i] * ff_pow_0_75[i];
282
        lpc_n[i] = lpc[i] * ff_pow_0_5 [i];
283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366
    };

    memcpy(pole_out - LP_FILTER_ORDER, ctx->postfilter_mem,
           LP_FILTER_ORDER*sizeof(float));

    ff_celp_lp_synthesis_filterf(pole_out, lpc_d, samples, SUBFR_SIZE,
                                 LP_FILTER_ORDER);

    memcpy(ctx->postfilter_mem, pole_out + SUBFR_SIZE - LP_FILTER_ORDER,
           LP_FILTER_ORDER*sizeof(float));

    ff_tilt_compensation(&ctx->tilt_mem, 0.4, pole_out, SUBFR_SIZE);

    memcpy(pole_out - LP_FILTER_ORDER, ctx->postfilter_mem5k0,
           LP_FILTER_ORDER*sizeof(*pole_out));

    memcpy(ctx->postfilter_mem5k0, pole_out + SUBFR_SIZE - LP_FILTER_ORDER,
           LP_FILTER_ORDER*sizeof(*pole_out));

    ff_celp_lp_zero_synthesis_filterf(samples, lpc_n, pole_out, SUBFR_SIZE,
                                      LP_FILTER_ORDER);

}

static void decode_fixed_sparse(AMRFixed *fixed_sparse, const int16_t *pulses,
                                SiprMode mode, int low_gain)
{
    int i;

    switch (mode) {
    case MODE_6k5:
        for (i = 0; i < 3; i++) {
            fixed_sparse->x[i] = 3 * (pulses[i] & 0xf) + i;
            fixed_sparse->y[i] = pulses[i] & 0x10 ? -1 : 1;
        }
        fixed_sparse->n = 3;
        break;
    case MODE_8k5:
        for (i = 0; i < 3; i++) {
            fixed_sparse->x[2*i    ] = 3 * ((pulses[i] >> 4) & 0xf) + i;
            fixed_sparse->x[2*i + 1] = 3 * ( pulses[i]       & 0xf) + i;

            fixed_sparse->y[2*i    ] = (pulses[i] & 0x100) ? -1.0: 1.0;

            fixed_sparse->y[2*i + 1] =
                (fixed_sparse->x[2*i + 1] < fixed_sparse->x[2*i]) ?
                -fixed_sparse->y[2*i    ] : fixed_sparse->y[2*i];
        }

        fixed_sparse->n = 6;
        break;
    case MODE_5k0:
    default:
        if (low_gain) {
            int offset = (pulses[0] & 0x200) ? 2 : 0;
            int val = pulses[0];

            for (i = 0; i < 3; i++) {
                int index = (val & 0x7) * 6 + 4 - i*2;

                fixed_sparse->y[i] = (offset + index) & 0x3 ? -1 : 1;
                fixed_sparse->x[i] = index;

                val >>= 3;
            }
            fixed_sparse->n = 3;
        } else {
            int pulse_subset = (pulses[0] >> 8) & 1;

            fixed_sparse->x[0] = ((pulses[0] >> 4) & 15) * 3 + pulse_subset;
            fixed_sparse->x[1] = ( pulses[0]       & 15) * 3 + pulse_subset + 1;

            fixed_sparse->y[0] = pulses[0] & 0x200 ? -1 : 1;
            fixed_sparse->y[1] = -fixed_sparse->y[0];
            fixed_sparse->n = 2;
        }
        break;
    }
}

static void decode_frame(SiprContext *ctx, SiprParameters *params,
                         float *out_data)
{
    int i, j;
367 368
    int subframe_count = modes[ctx->mode].subframe_count;
    int frame_size = subframe_count * SUBFR_SIZE;
369
    float Az[LP_FILTER_ORDER * MAX_SUBFRAME_COUNT];
370 371 372 373 374 375 376 377 378 379 380 381
    float *excitation;
    float ir_buf[SUBFR_SIZE + LP_FILTER_ORDER];
    float lsf_new[LP_FILTER_ORDER];
    float *impulse_response = ir_buf + LP_FILTER_ORDER;
    float *synth = ctx->synth_buf + 16; // 16 instead of LP_FILTER_ORDER for
                                        // memory alignment
    int t0_first = 0;
    AMRFixed fixed_cb;

    memset(ir_buf, 0, LP_FILTER_ORDER * sizeof(float));
    lsf_decode_fp(lsf_new, ctx->lsf_history, params);

382
    sipr_decode_lp(lsf_new, ctx->lsp_history, Az, subframe_count);
383 384 385 386 387

    memcpy(ctx->lsp_history, lsf_new, LP_FILTER_ORDER * sizeof(float));

    excitation = ctx->excitation + PITCH_DELAY_MAX + L_INTERPOL;

388
    for (i = 0; i < subframe_count; i++) {
389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407
        float *pAz = Az + i*LP_FILTER_ORDER;
        float fixed_vector[SUBFR_SIZE];
        int T0,T0_frac;
        float pitch_gain, gain_code, avg_energy;

        ff_decode_pitch_lag(&T0, &T0_frac, params->pitch_delay[i], t0_first, i,
                            ctx->mode == MODE_5k0, 6);

        if (i == 0 || (i == 2 && ctx->mode == MODE_5k0))
            t0_first = T0;

        ff_acelp_interpolatef(excitation, excitation - T0 + (T0_frac <= 0),
                              ff_b60_sinc, 6,
                              2 * ((2 + T0_frac)%3 + 1), LP_FILTER_ORDER,
                              SUBFR_SIZE);

        decode_fixed_sparse(&fixed_cb, params->fc_indexes[i], ctx->mode,
                            ctx->past_pitch_gain < 0.8);

408
        eval_ir(pAz, T0, impulse_response, modes[ctx->mode].pitch_sharp_factor);
409 410 411 412 413 414 415 416 417 418 419 420

        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;

        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,
421
                                          34 - 15.0/(0.05*M_LN10/M_LN2),
422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454
                                          pred);

        ff_weighted_vector_sumf(excitation, excitation, fixed_vector,
                                pitch_gain, gain_code, SUBFR_SIZE);

        pitch_gain *= 0.5 * pitch_gain;
        pitch_gain = FFMIN(pitch_gain, 0.4);

        ctx->gain_mem = 0.7 * ctx->gain_mem + 0.3 * pitch_gain;
        ctx->gain_mem = FFMIN(ctx->gain_mem, pitch_gain);
        gain_code *= ctx->gain_mem;

        for (j = 0; j < SUBFR_SIZE; j++)
            fixed_vector[j] = excitation[j] - gain_code * fixed_vector[j];

        if (ctx->mode == MODE_5k0) {
            postfilter_5k0(ctx, pAz, fixed_vector);

            ff_celp_lp_synthesis_filterf(ctx->postfilter_syn5k0 + LP_FILTER_ORDER + i*SUBFR_SIZE,
                                         pAz, excitation, SUBFR_SIZE,
                                         LP_FILTER_ORDER);
        }

        ff_celp_lp_synthesis_filterf(synth + i*SUBFR_SIZE, pAz, fixed_vector,
                                     SUBFR_SIZE, LP_FILTER_ORDER);

        excitation += SUBFR_SIZE;
    }

    memcpy(synth - LP_FILTER_ORDER, synth + frame_size - LP_FILTER_ORDER,
           LP_FILTER_ORDER * sizeof(float));

    if (ctx->mode == MODE_5k0) {
455
        for (i = 0; i < subframe_count; i++) {
456 457 458
            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);
459 460
            ff_adaptive_gain_control(&synth[i * SUBFR_SIZE],
                                     &synth[i * SUBFR_SIZE], energy,
Ronald S. Bultje's avatar
Ronald S. Bultje committed
461
                                     SUBFR_SIZE, 0.9, &ctx->postfilter_agc);
462 463 464 465 466
        }

        memcpy(ctx->postfilter_syn5k0, ctx->postfilter_syn5k0 + frame_size,
               LP_FILTER_ORDER*sizeof(float));
    }
467
    memmove(ctx->excitation, excitation - PITCH_DELAY_MAX - L_INTERPOL,
468 469
           (PITCH_DELAY_MAX + L_INTERPOL) * sizeof(float));

470
    ff_acelp_apply_order_2_transfer_function(out_data, synth,
471 472 473 474 475 476 477 478 479 480 481 482
                                             (const float[2]) {-1.99997   , 1.000000000},
                                             (const float[2]) {-1.93307352, 0.935891986},
                                             0.939805806,
                                             ctx->highpass_filt_mem,
                                             frame_size);
}

static av_cold int sipr_decoder_init(AVCodecContext * avctx)
{
    SiprContext *ctx = avctx->priv_data;
    int i;

483 484 485 486 487 488 489 490 491
    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:
        av_log(avctx, AV_LOG_ERROR, "Invalid block_align: %d\n", avctx->block_align);
        return AVERROR(EINVAL);
    }
492

493
    av_log(avctx, AV_LOG_DEBUG, "Mode: %s\n", modes[ctx->mode].mode_name);
494

495
    if (ctx->mode == MODE_16k) {
Vitor Sessak's avatar
Vitor Sessak committed
496
        ff_sipr_init_16k(ctx);
497 498 499 500
        ctx->decode_frame = ff_sipr_decode_frame_16k;
    } else {
        ctx->decode_frame = decode_frame;
    }
Vitor Sessak's avatar
Vitor Sessak committed
501

502 503 504 505 506 507
    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;

508
    avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
509

510 511 512
    avcodec_get_frame_defaults(&ctx->frame);
    avctx->coded_frame = &ctx->frame;

513 514 515
    return 0;
}

516 517
static int sipr_decode_frame(AVCodecContext *avctx, void *data,
                             int *got_frame_ptr, AVPacket *avpkt)
518 519 520 521
{
    SiprContext *ctx = avctx->priv_data;
    const uint8_t *buf=avpkt->data;
    SiprParameters parm;
522
    const SiprModeParam *mode_par = &modes[ctx->mode];
523
    GetBitContext gb;
524
    float *samples;
Vitor Sessak's avatar
Vitor Sessak committed
525
    int subframe_size = ctx->mode == MODE_16k ? L_SUBFR_16k : SUBFR_SIZE;
526
    int i, ret;
527 528

    ctx->avctx = avctx;
529
    if (avpkt->size < (mode_par->bits_per_frame >> 3)) {
530 531 532 533 534
        av_log(avctx, AV_LOG_ERROR,
               "Error processing packet: packet size (%d) too small\n",
               avpkt->size);
        return -1;
    }
535

536 537 538 539 540 541
    /* get output buffer */
    ctx->frame.nb_samples = mode_par->frames_per_packet * subframe_size *
                            mode_par->subframe_count;
    if ((ret = avctx->get_buffer(avctx, &ctx->frame)) < 0) {
        av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
        return ret;
542
    }
543
    samples = (float *)ctx->frame.data[0];
544

545
    init_get_bits(&gb, buf, mode_par->bits_per_frame);
546

547 548
    for (i = 0; i < mode_par->frames_per_packet; i++) {
        decode_parameters(&parm, &gb, mode_par);
549

550
        ctx->decode_frame(ctx, &parm, samples);
Vitor Sessak's avatar
Vitor Sessak committed
551

552
        samples += subframe_size * mode_par->subframe_count;
553 554
    }

555 556
    *got_frame_ptr   = 1;
    *(AVFrame *)data = ctx->frame;
557

558
    return mode_par->bits_per_frame >> 3;
559
}
560

561
AVCodec ff_sipr_decoder = {
562 563 564 565 566 567
    .name           = "sipr",
    .type           = AVMEDIA_TYPE_AUDIO,
    .id             = CODEC_ID_SIPR,
    .priv_data_size = sizeof(SiprContext),
    .init           = sipr_decoder_init,
    .decode         = sipr_decode_frame,
568
    .capabilities   = CODEC_CAP_DR1,
569
    .long_name      = NULL_IF_CONFIG_SMALL("RealAudio SIPR / ACELP.NET"),
570
};