41 #define AMR_USE_16BIT_TABLES
115 ctx->first_frame = 1;
120 for (
i = 0;
i < 4;
i++)
143 ctx->fr_cur_mode = buf[0] >> 3 & 0x0F;
144 ctx->fr_quality = (buf[0] & 0x4) == 0x4;
159 for (
i = 0;
i < 9;
i++)
160 isf_q[
i] =
dico1_isf[ind[0]][
i] * (1.0f / (1 << 15));
162 for (
i = 0;
i < 7;
i++)
163 isf_q[
i + 9] =
dico2_isf[ind[1]][
i] * (1.0f / (1 << 15));
165 for (
i = 0;
i < 5;
i++)
168 for (
i = 0;
i < 4;
i++)
171 for (
i = 0;
i < 7;
i++)
185 for (
i = 0;
i < 9;
i++)
186 isf_q[
i] =
dico1_isf[ind[0]][
i] * (1.0f / (1 << 15));
188 for (
i = 0;
i < 7;
i++)
189 isf_q[
i + 9] =
dico2_isf[ind[1]][
i] * (1.0f / (1 << 15));
191 for (
i = 0;
i < 3;
i++)
194 for (
i = 0;
i < 3;
i++)
195 isf_q[
i + 3] +=
dico22_isf[ind[3]][
i] * (1.0f / (1 << 15));
197 for (
i = 0;
i < 3;
i++)
198 isf_q[
i + 6] +=
dico23_isf[ind[4]][
i] * (1.0f / (1 << 15));
200 for (
i = 0;
i < 3;
i++)
201 isf_q[
i + 9] +=
dico24_isf[ind[5]][
i] * (1.0f / (1 << 15));
203 for (
i = 0;
i < 4;
i++)
204 isf_q[
i + 12] +=
dico25_isf[ind[6]][
i] * (1.0f / (1 << 15));
238 for (k = 0; k < 3; k++) {
241 isp_q[k][
i] = (1.0 -
c) * isp4_past[
i] +
c * isp_q[3][
i];
257 uint8_t *base_lag_int,
int subframe)
259 if (subframe == 0 || subframe == 2) {
260 if (pitch_index < 376) {
261 *lag_int = (pitch_index + 137) >> 2;
262 *lag_frac = pitch_index - (*lag_int << 2) + 136;
263 }
else if (pitch_index < 440) {
264 *lag_int = (pitch_index + 257 - 376) >> 1;
265 *lag_frac = (pitch_index - (*lag_int << 1) + 256 - 376) * 2;
268 *lag_int = pitch_index - 280;
272 *base_lag_int =
av_clip(*lag_int - 8 - (*lag_frac < 0),
278 *lag_int = (pitch_index + 1) >> 2;
279 *lag_frac = pitch_index - (*lag_int << 2);
280 *lag_int += *base_lag_int;
293 if (pitch_index < 116) {
294 *lag_int = (pitch_index + 69) >> 1;
295 *lag_frac = (pitch_index - (*lag_int << 1) + 68) * 2;
297 *lag_int = pitch_index - 24;
301 *base_lag_int =
av_clip(*lag_int - 8 - (*lag_frac < 0),
304 *lag_int = (pitch_index + 1) >> 1;
305 *lag_frac = (pitch_index - (*lag_int << 1)) * 2;
306 *lag_int += *base_lag_int;
322 int pitch_lag_int, pitch_lag_frac;
324 float *exc =
ctx->excitation;
329 &
ctx->base_pitch_lag, subframe,
mode);
332 &
ctx->base_pitch_lag, subframe);
334 ctx->pitch_lag_int = pitch_lag_int;
335 pitch_lag_int += pitch_lag_frac > 0;
339 ctx->acelpf_ctx.acelp_interpolatef(exc,
340 exc + 1 - pitch_lag_int,
342 pitch_lag_frac + (pitch_lag_frac > 0 ? 0 : 4),
347 if (amr_subframe->
ltp) {
351 ctx->pitch_vector[
i] = 0.18 * exc[
i - 1] + 0.64 * exc[
i] +
358 #define BIT_STR(x,lsb,len) av_mod_uintp2((x) >> (lsb), (len))
361 #define BIT_POS(x, p) (((x) >> (p)) & 1)
398 m - 1, off + half_2p);
404 int half_4p, subhalf_2p;
405 int b_offset = 1 << (m - 1);
413 m - 2, off + half_4p + subhalf_2p);
415 m - 1, off + half_4p);
421 m - 1, off + b_offset);
427 m - 1, off + b_offset);
433 m - 1, off + b_offset);
443 m - 1, off + half_3p);
450 int b_offset = 1 << (m - 1);
453 int half_other = b_offset - half_more;
458 m - 1, off + half_more);
460 m - 1, off + half_more);
464 m - 1, off + half_other);
466 m - 1, off + half_more);
470 m - 1, off + half_other);
472 m - 1, off + half_more);
478 m - 1, off + b_offset);
493 const uint16_t *pulse_lo,
const enum Mode mode)
503 for (
i = 0;
i < 2;
i++)
507 for (
i = 0;
i < 4;
i++)
511 for (
i = 0;
i < 4;
i++)
515 for (
i = 0;
i < 2;
i++)
517 for (
i = 2;
i < 4;
i++)
521 for (
i = 0;
i < 4;
i++)
525 for (
i = 0;
i < 4;
i++)
527 ((
int) pulse_hi[
i] << 14), 4, 1);
530 for (
i = 0;
i < 2;
i++)
532 ((
int) pulse_hi[
i] << 10), 4, 1);
533 for (
i = 2;
i < 4;
i++)
535 ((
int) pulse_hi[
i] << 14), 4, 1);
539 for (
i = 0;
i < 4;
i++)
541 ((
int) pulse_hi[
i] << 11), 4, 1);
547 for (
i = 0;
i < 4;
i++)
549 int pos = (
FFABS(sig_pos[
i][j]) - 1) * spacing +
i;
551 fixed_vector[
pos] += sig_pos[
i][j] < 0 ? -1.0 : 1.0;
564 float *fixed_gain_factor,
float *pitch_gain)
569 *pitch_gain = gains[0] * (1.0f / (1 << 14));
570 *fixed_gain_factor = gains[1] * (1.0f / (1 << 11));
587 fixed_vector[
i] -= fixed_vector[
i - 1] *
ctx->tilt_coef;
591 fixed_vector[
i] += fixed_vector[
i -
ctx->pitch_lag_int] * 0.85;
604 float *f_vector,
float f_gain,
607 double p_ener = (double)
ctx->dot_productf(p_vector, p_vector,
610 double f_ener = (double)
ctx->dot_productf(f_vector, f_vector,
614 return (p_ener - f_ener) / (p_ener + f_ener + 0.01);
628 float *fixed_vector,
float *buf)
635 if (
ctx->pitch_gain[0] < 0.6) {
637 }
else if (
ctx->pitch_gain[0] < 0.9) {
643 if (
ctx->fixed_gain[0] > 3.0 *
ctx->fixed_gain[1]) {
644 if (ir_filter_nr < 2)
649 for (
i = 0;
i < 6;
i++)
650 if (
ctx->pitch_gain[
i] < 0.6)
656 if (ir_filter_nr >
ctx->prev_ir_filter_nr + 1)
661 ctx->prev_ir_filter_nr = ir_filter_nr;
665 if (ir_filter_nr < 2) {
697 acc += (isf[
i] - isf_past[
i]) * (isf[
i] - isf_past[
i]);
701 return FFMAX(0.0, 1.25 -
acc * 0.8 * 512);
716 float voice_fac,
float stab_fac)
718 float sm_fac = 0.5 * (1 - voice_fac) * stab_fac;
724 if (fixed_gain < *prev_tr_gain) {
725 g0 =
FFMIN(*prev_tr_gain, fixed_gain + fixed_gain *
726 (6226 * (1.0f / (1 << 15))));
728 g0 =
FFMAX(*prev_tr_gain, fixed_gain *
729 (27536 * (1.0f / (1 << 15))));
733 return sm_fac * g0 + (1 - sm_fac) * fixed_gain;
745 float cpe = 0.125 * (1 + voice_fac);
746 float last = fixed_vector[0];
748 fixed_vector[0] -= cpe * fixed_vector[1];
751 float cur = fixed_vector[
i];
753 fixed_vector[
i] -= cpe * (last + fixed_vector[
i + 1]);
771 float fixed_gain,
const float *fixed_vector,
774 ctx->acelpv_ctx.weighted_vector_sumf(excitation,
ctx->pitch_vector, fixed_vector,
780 float energy =
ctx->celpm_ctx.dot_productf(excitation, excitation,
785 float pitch_factor = 0.25 *
ctx->pitch_gain[0] *
ctx->pitch_gain[0];
788 excitation[
i] += pitch_factor *
ctx->pitch_vector[
i];
794 ctx->celpf_ctx.celp_lp_synthesis_filterf(samples, lpc, excitation,
811 out[0] =
in[0] + m * mem[0];
832 int int_part = 0, frac_part;
835 for (j = 0; j < o_size / 5; j++) {
840 for (k = 1; k < 5; k++) {
841 out[
i] =
ctx->dot_productf(in0 + int_part,
878 return av_clipf((1.0 - tilt) * (1.25 - 0.25 * wsp), 0.1, 1.0);
891 const float *synth_exc,
float hb_gain)
894 float energy =
ctx->celpm_ctx.dot_productf(synth_exc, synth_exc,
902 energy * hb_gain * hb_gain,
915 float prod = (diff_isf[
i] -
mean) * (diff_isf[
i - lag] -
mean);
930 float diff_isf[
LP_ORDER - 2], diff_mean;
933 int i, j, i_max_corr;
939 diff_isf[
i] = isf[
i + 1] - isf[
i];
943 diff_mean += diff_isf[
i] * (1.0f / (
LP_ORDER - 4));
947 for (
i = 0;
i < 3;
i++) {
950 if (corr_lag[
i] > corr_lag[i_max_corr])
956 isf[
i] = isf[
i - 1] + isf[
i - 1 - i_max_corr]
957 - isf[
i - 2 - i_max_corr];
960 est = 7965 + (isf[2] - isf[3] - isf[4]) / 6.0;
965 diff_isf[j] = scale * (isf[
i] - isf[
i - 1]);
969 if (diff_isf[
i] + diff_isf[
i - 1] < 5.0) {
970 if (diff_isf[
i] > diff_isf[
i - 1]) {
971 diff_isf[
i - 1] = 5.0 - diff_isf[
i];
973 diff_isf[
i] = 5.0 - diff_isf[
i - 1];
977 isf[
i] = isf[
i - 1] + diff_isf[j] * (1.0f / (1 << 15));
999 out[
i] = lpc[
i] * fac;
1016 const float *exc,
const float *isf,
const float *isf_past)
1025 ctx->acelpv_ctx.weighted_vector_sumf(e_isf, isf_past, isf,
isfp_inter[subframe],
1055 #ifndef hb_fir_filter
1083 memmove(&
ctx->pitch_gain[1], &
ctx->pitch_gain[0], 5 *
sizeof(
float));
1084 memmove(&
ctx->fixed_gain[1], &
ctx->fixed_gain[0], 1 *
sizeof(
float));
1095 int *got_frame_ptr,
AVPacket *avpkt)
1101 int buf_size = avpkt->
size;
1102 int expected_fr_size, header_size;
1105 float fixed_gain_factor;
1106 float *synth_fixed_vector;
1107 float synth_fixed_gain;
1108 float voice_fac, stab_fac;
1122 expected_fr_size = ((
cf_sizes_wb[
ctx->fr_cur_mode] + 7) >> 3) + 1;
1124 if (!
ctx->fr_quality)
1132 return expected_fr_size;
1136 "Invalid mode %d\n",
ctx->fr_cur_mode);
1140 if (buf_size < expected_fr_size) {
1142 "Frame too small (%d bytes). Truncated file?\n", buf_size);
1171 if (
ctx->first_frame) {
1172 ctx->first_frame = 0;
1173 memcpy(
ctx->isp_sub4_past,
ctx->isp[3],
LP_ORDER *
sizeof(
double));
1188 cur_subframe->
pul_il,
ctx->fr_cur_mode);
1193 &fixed_gain_factor, &
ctx->pitch_gain[0]);
1195 ctx->fixed_gain[0] =
1197 ctx->celpm_ctx.dot_productf(
ctx->fixed_vector,
1201 ctx->prediction_error,
1206 ctx->fixed_vector,
ctx->fixed_gain[0],
1208 ctx->tilt_coef = voice_fac * 0.25 + 0.25;
1212 ctx->excitation[
i] *=
ctx->pitch_gain[0];
1213 ctx->excitation[
i] +=
ctx->fixed_gain[0] *
ctx->fixed_vector[
i];
1219 voice_fac, stab_fac);
1233 ctx->acelpf_ctx.acelp_apply_order_2_transfer_function(&
ctx->samples_up[
UPS_MEM_SIZE],
1241 ctx->acelpf_ctx.acelp_apply_order_2_transfer_function(hb_samples,
1251 hb_exc,
ctx->isf_cur,
ctx->isf_past_final);
1263 sub_buf[
i] = (sub_buf[
i] + hb_samples[
i]) * (1.0f / (1 << 15));
1271 memcpy(
ctx->isf_past_final,
ctx->isf_cur,
LP_ORDER *
sizeof(
float));
1275 return expected_fr_size;
void ff_acelp_filter_init(ACELPFContext *c)
Initialize ACELPFContext.
float ff_amr_set_fixed_gain(float fixed_gain_factor, float fixed_mean_energy, float *prediction_error, float energy_mean, const float *pred_table)
Calculate fixed gain (part of section 6.1.3 of AMR spec)
void ff_acelp_vectors_init(ACELPVContext *c)
Initialize ACELPVContext.
void ff_scale_vector_to_given_sum_of_squares(float *out, const float *in, float sum_of_squares, const int n)
Set the sum of squares of a signal by scaling.
static enum AVSampleFormat sample_fmts[]
static void ff_amr_bit_reorder(uint16_t *out, int size, const uint8_t *data, const R_TABLE_TYPE *ord_table)
Fill the frame structure variables from bitstream by parsing the given reordering table that uses the...
static const float *const ir_filters_lookup[2]
Mode
Frame type (Table 1a in 3GPP TS 26.101)
static const float energy_pred_fac[4]
4-tap moving average prediction coefficients in reverse order
#define MIN_ENERGY
Initial energy in dB.
AMR wideband data and definitions.
static const int16_t dico24_isf[32][3]
static const int16_t qua_gain_7b[128][2]
static const uint16_t qua_hb_gain[16]
High band quantized gains for 23k85 in Q14.
#define MIN_ISF_SPACING
minimum isf gap
static const float hpf_zeros[2]
High-pass filters coefficients for 31 Hz and 400 Hz cutoff.
static const float lpf_7_coef[31]
static const int16_t dico1_isf[256][9]
Indexed tables for retrieval of quantized ISF vectors in Q15.
#define HB_FIR_SIZE
amount of past data needed by HB filters
static const uint16_t *const amr_bit_orderings_by_mode[]
Reordering array addresses for each mode.
static const int16_t isf_mean[LP_ORDER]
Means of ISF vectors in Q15.
static const float hpf_31_poles[2]
@ MODE_SID
comfort noise frame
static const int16_t dico25_isf[32][4]
#define AMRWB_P_DELAY_MIN
static const int16_t qua_gain_6b[64][2]
Tables for decoding quantized gains { pitch (Q14), fixed factor (Q11) }.
#define AMRWB_P_DELAY_MAX
maximum pitch delay value
static const float hpf_400_gain
static const int16_t dico21_isf[64][3]
static const int16_t isf_init[LP_ORDER]
Initialization tables for the processed ISF vector in Q15.
#define UPS_FIR_SIZE
upsampling filter size
static const float hpf_31_gain
static const int16_t dico23_isf_36b[64][7]
static const int16_t dico2_isf[256][7]
static const uint16_t cf_sizes_wb[]
Core frame sizes in each mode.
static const float ac_inter[65]
Coefficients for FIR interpolation of excitation vector at pitch lag resulting the adaptive codebook ...
static const float hpf_400_poles[2]
static const float bpf_6_7_coef[31]
High-band post-processing FIR filters coefficients from Q15.
#define LP_ORDER
linear predictive coding filter order
#define PREEMPH_FAC
factor used to de-emphasize synthesis
static const float upsample_fir[4][24]
Interpolation coefficients for 5/4 signal upsampling Table from the reference source was reordered fo...
#define ENERGY_MEAN
mean innovation energy (dB) in all modes
static const int16_t dico21_isf_36b[128][5]
#define AMRWB_SFR_SIZE_16k
samples per subframe at 16 kHz
static const int16_t dico22_isf[128][3]
static const int16_t dico23_isf[128][3]
static const float isfp_inter[4]
ISF/ISP interpolation coefficients for each subframe.
#define LP_ORDER_16k
lpc filter order at 16kHz
static const int16_t dico22_isf_36b[128][4]
static const uint8_t pulses_nb_per_mode_tr[][4]
[i][j] is the number of pulses present in track j at mode i
#define AMRWB_SFR_SIZE
samples per subframe at 12.8 kHz
static void decode_gains(const uint8_t vq_gain, const enum Mode mode, float *fixed_gain_factor, float *pitch_gain)
Decode pitch gain and fixed gain correction factor.
static void extrapolate_isf(float isf[LP_ORDER_16k])
Extrapolate a ISF vector to the 16kHz range (20th order LP) used at mode 6k60 LP filter for the high ...
static void update_sub_state(AMRWBContext *ctx)
Update context state before the next subframe.
static void decode_6p_track(int *out, int code, int m, int off)
code: 6m-2 bits
static void decode_pitch_vector(AMRWBContext *ctx, const AMRWBSubFrame *amr_subframe, const int subframe)
Find the pitch vector by interpolating the past excitation at the pitch delay, which is obtained in t...
static void decode_1p_track(int *out, int code, int m, int off)
The next six functions decode_[i]p_track decode exactly i pulses positions and amplitudes (-1 or 1) i...
static void isf_add_mean_and_past(float *isf_q, float *isf_past)
Apply mean and past ISF values using the prediction factor.
static void upsample_5_4(float *out, const float *in, int o_size, CELPMContext *ctx)
Upsample a signal by 5/4 ratio (from 12.8kHz to 16kHz) using a FIR interpolation filter.
static void scaled_hb_excitation(AMRWBContext *ctx, float *hb_exc, const float *synth_exc, float hb_gain)
Generate the high-band excitation with the same energy from the lower one and scaled by the given gai...
static float stability_factor(const float *isf, const float *isf_past)
Calculate a stability factor {teta} based on distance between current and past isf.
static void decode_pitch_lag_low(int *lag_int, int *lag_frac, int pitch_index, uint8_t *base_lag_int, int subframe, enum Mode mode)
Decode an adaptive codebook index into pitch lag for 8k85 and 6k60 modes.
static float voice_factor(float *p_vector, float p_gain, float *f_vector, float f_gain, CELPMContext *ctx)
Calculate the voicing factor (-1.0 = unvoiced to 1.0 = voiced).
static void decode_isf_indices_36b(uint16_t *ind, float *isf_q)
Decode quantized ISF vectors using 36-bit indexes (6K60 mode only).
static void pitch_enhancer(float *fixed_vector, float voice_fac)
Filter the fixed_vector to emphasize the higher frequencies.
static void decode_fixed_vector(float *fixed_vector, const uint16_t *pulse_hi, const uint16_t *pulse_lo, const enum Mode mode)
Decode the algebraic codebook index to pulse positions and signs, then construct the algebraic codebo...
static void interpolate_isp(double isp_q[4][LP_ORDER], const double *isp4_past)
Interpolate the fourth ISP vector from current and past frames to obtain an ISP vector for each subfr...
static void decode_pitch_lag_high(int *lag_int, int *lag_frac, int pitch_index, uint8_t *base_lag_int, int subframe)
Decode an adaptive codebook index into pitch lag (except 6k60, 8k85 modes).
static int amrwb_decode_frame(AVCodecContext *avctx, void *data, int *got_frame_ptr, AVPacket *avpkt)
static void hb_synthesis(AMRWBContext *ctx, int subframe, float *samples, const float *exc, const float *isf, const float *isf_past)
Conduct 20th order linear predictive coding synthesis for the high frequency band excitation at 16kHz...
static float * anti_sparseness(AMRWBContext *ctx, float *fixed_vector, float *buf)
Reduce fixed vector sparseness by smoothing with one of three IR filters, also known as "adaptive pha...
static av_cold int amrwb_decode_init(AVCodecContext *avctx)
static float noise_enhancer(float fixed_gain, float *prev_tr_gain, float voice_fac, float stab_fac)
Apply a non-linear fixed gain smoothing in order to reduce fluctuation in the energy of excitation.
#define BIT_POS(x, p)
Get the bit at specified position.
#define BIT_STR(x, lsb, len)
Get x bits in the index interval [lsb,lsb+len-1] inclusive.
static int decode_mime_header(AMRWBContext *ctx, const uint8_t *buf)
Decode the frame header in the "MIME/storage" format.
static void decode_4p_track(int *out, int code, int m, int off)
code: 4m bits
static void hb_fir_filter(float *out, const float fir_coef[HB_FIR_SIZE+1], float mem[HB_FIR_SIZE], const float *in)
Apply a 15th order filter to high-band samples.
static float auto_correlation(float *diff_isf, float mean, int lag)
Calculate the auto-correlation for the ISF difference vector.
static void lpc_weighting(float *out, const float *lpc, float gamma, int size)
Spectral expand the LP coefficients using the equation: y[i] = x[i] * (gamma ** i)
static void decode_isf_indices_46b(uint16_t *ind, float *isf_q)
Decode quantized ISF vectors using 46-bit indexes (except 6K60 mode).
static void de_emphasis(float *out, float *in, float m, float mem[1])
Apply to synthesis a de-emphasis filter of the form: H(z) = 1 / (1 - m * z^-1)
static void synthesis(AMRWBContext *ctx, float *lpc, float *excitation, float fixed_gain, const float *fixed_vector, float *samples)
Conduct 16th order linear predictive coding synthesis from excitation.
static void decode_5p_track(int *out, int code, int m, int off)
code: 5m bits
static float find_hb_gain(AMRWBContext *ctx, const float *synth, uint16_t hb_idx, uint8_t vad)
Calculate the high-band gain based on encoded index (23k85 mode) or on the low-band speech signal and...
static void pitch_sharpening(AMRWBContext *ctx, float *fixed_vector)
Apply pitch sharpening filters to the fixed codebook vector.
static void decode_3p_track(int *out, int code, int m, int off)
code: 3m+1 bits
static void decode_2p_track(int *out, int code, int m, int off)
code: 2m+1 bits
Reference: libavcodec/amrwbdec.c.
uint8_t pi<< 24) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_U8, uint8_t,(*(const uint8_t *) pi - 0x80) *(1.0f/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_U8, uint8_t,(*(const uint8_t *) pi - 0x80) *(1.0/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S16, int16_t,(*(const int16_t *) pi >> 8)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S16, int16_t, *(const int16_t *) pi *(1.0f/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S16, int16_t, *(const int16_t *) pi *(1.0/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S32, int32_t,(*(const int32_t *) pi >> 24)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S32, int32_t, *(const int32_t *) pi *(1.0f/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S32, int32_t, *(const int32_t *) pi *(1.0/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_FLT, float, av_clip_uint8(lrintf(*(const float *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_FLT, float, av_clip_int16(lrintf(*(const float *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_FLT, float, av_clipl_int32(llrintf(*(const float *) pi *(1U<< 31)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_DBL, double, av_clip_uint8(lrint(*(const double *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_DBL, double, av_clip_int16(lrint(*(const double *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_DBL, double, av_clipl_int32(llrint(*(const double *) pi *(1U<< 31)))) #define SET_CONV_FUNC_GROUP(ofmt, ifmt) static void set_generic_function(AudioConvert *ac) { } void ff_audio_convert_free(AudioConvert **ac) { if(! *ac) return;ff_dither_free(&(*ac) ->dc);av_freep(ac);} AudioConvert *ff_audio_convert_alloc(AVAudioResampleContext *avr, enum AVSampleFormat out_fmt, enum AVSampleFormat in_fmt, int channels, int sample_rate, int apply_map) { AudioConvert *ac;int in_planar, out_planar;ac=av_mallocz(sizeof(*ac));if(!ac) return NULL;ac->avr=avr;ac->out_fmt=out_fmt;ac->in_fmt=in_fmt;ac->channels=channels;ac->apply_map=apply_map;if(avr->dither_method !=AV_RESAMPLE_DITHER_NONE &&av_get_packed_sample_fmt(out_fmt)==AV_SAMPLE_FMT_S16 &&av_get_bytes_per_sample(in_fmt) > 2) { ac->dc=ff_dither_alloc(avr, out_fmt, in_fmt, channels, sample_rate, apply_map);if(!ac->dc) { av_free(ac);return NULL;} return ac;} in_planar=ff_sample_fmt_is_planar(in_fmt, channels);out_planar=ff_sample_fmt_is_planar(out_fmt, channels);if(in_planar==out_planar) { ac->func_type=CONV_FUNC_TYPE_FLAT;ac->planes=in_planar ? ac->channels :1;} else if(in_planar) ac->func_type=CONV_FUNC_TYPE_INTERLEAVE;else ac->func_type=CONV_FUNC_TYPE_DEINTERLEAVE;set_generic_function(ac);if(ARCH_AARCH64) ff_audio_convert_init_aarch64(ac);if(ARCH_ARM) ff_audio_convert_init_arm(ac);if(ARCH_X86) ff_audio_convert_init_x86(ac);return ac;} int ff_audio_convert(AudioConvert *ac, AudioData *out, AudioData *in) { int use_generic=1;int len=in->nb_samples;int p;if(ac->dc) { av_log(ac->avr, AV_LOG_TRACE, "%d samples - audio_convert: %s to %s (dithered)\n", len, av_get_sample_fmt_name(ac->in_fmt), av_get_sample_fmt_name(ac->out_fmt));return ff_convert_dither(ac-> in
Libavcodec external API header.
static av_cold int init(AVCodecContext *avctx)
void ff_celp_circ_addf(float *out, const float *in, const float *lagged, int lag, float fac, int n)
Add an array to a rotated array.
void ff_celp_filter_init(CELPFContext *c)
Initialize CELPFContext.
void ff_celp_math_init(CELPMContext *c)
Initialize CELPMContext.
audio channel layout utility functions
common internal and external API header
#define FFABS(a)
Absolute value, Note, INT_MIN / INT64_MIN result in undefined behavior as they are not representable ...
static __device__ float truncf(float a)
int ff_get_buffer(AVCodecContext *avctx, AVFrame *frame, int flags)
Get a buffer for a frame.
static void decode(AVCodecContext *dec_ctx, AVPacket *pkt, AVFrame *frame, FILE *outfile)
static float sub(float src0, float src1)
mode
Use these values in ebur128_init (or'ed).
#define AV_CH_LAYOUT_MONO
#define AV_CODEC_CAP_DR1
Codec uses get_buffer() or get_encode_buffer() for allocating buffers and supports custom allocators.
#define AV_CODEC_CAP_CHANNEL_CONF
Codec should fill in channel configuration and samplerate instead of container.
#define AVERROR_PATCHWELCOME
Not yet implemented in FFmpeg, patches welcome.
#define AVERROR_INVALIDDATA
Invalid data found when processing input.
#define AV_LOG_ERROR
Something went wrong and cannot losslessly be recovered.
AVSampleFormat
Audio sample formats.
int av_samples_set_silence(uint8_t **audio_data, int offset, int nb_samples, int nb_channels, enum AVSampleFormat sample_fmt)
Fill an audio buffer with silence.
av_cold void av_lfg_init(AVLFG *c, unsigned int seed)
static unsigned int av_lfg_get(AVLFG *c)
Get the next random unsigned 32-bit number using an ALFG.
common internal API header
#define NULL_IF_CONFIG_SMALL(x)
Return NULL if CONFIG_SMALL is true, otherwise the argument without modification.
void avpriv_report_missing_feature(void *avc, const char *msg,...) av_printf_format(2
Log a generic warning message about a missing feature.
void ff_set_min_dist_lsf(float *lsf, double min_spacing, int size)
Adjust the quantized LSFs so they are increasing and not too close.
void ff_acelp_lsf2lspd(double *lsp, const float *lsf, int lp_order)
Floating point version of ff_acelp_lsf2lsp()
void ff_amrwb_lsp2lpc(const double *lsp, float *lp, int lp_order)
LSP to LP conversion (5.2.4 of AMR-WB)
CELPMContext celpm_ctx
context for fixed point math operations
uint8_t base_pitch_lag
integer part of pitch lag for the next relative subframe
uint8_t first_frame
flag active during decoding of the first frame
float prev_sparse_fixed_gain
previous fixed gain; used by anti-sparseness to determine "onset"
CELPFContext celpf_ctx
context for filters for CELP-based codecs
float isf_past_final[LP_ORDER]
final processed ISF vector of the previous frame
ACELPFContext acelpf_ctx
context for filters for ACELP-based codecs
AMRWBFrame frame
AMRWB parameters decoded from bitstream.
float samples_az[LP_ORDER+AMRWB_SFR_SIZE]
low-band samples and memory from synthesis at 12.8kHz
float pitch_gain[6]
quantified pitch gains for the current and previous five subframes
float isf_q_past[LP_ORDER]
quantized ISF vector of the previous frame
ACELPVContext acelpv_ctx
context for vector operations for ACELP-based codecs
float excitation_buf[AMRWB_P_DELAY_MAX+LP_ORDER+2+AMRWB_SFR_SIZE]
current excitation and all necessary excitation history
float fixed_vector[AMRWB_SFR_SIZE]
algebraic codebook (fixed) vector for current subframe
float hpf_400_mem[2]
previous values in the high pass filters
float bpf_6_7_mem[HB_FIR_SIZE]
previous values in the high-band band pass filter
AVLFG prng
random number generator for white noise excitation
float lpf_7_mem[HB_FIR_SIZE]
previous values in the high-band low pass filter
float * excitation
points to current excitation in excitation_buf[]
float prev_tr_gain
previous initial gain used by noise enhancer for threshold
uint8_t fr_quality
frame quality index (FQI)
float samples_hb[LP_ORDER_16k+AMRWB_SFR_SIZE_16k]
high-band samples and memory from synthesis at 16kHz
double isp[4][LP_ORDER]
ISP vectors from current frame.
uint8_t pitch_lag_int
integer part of pitch lag of the previous subframe
uint8_t prev_ir_filter_nr
previous impulse response filter "impNr": 0 - strong, 1 - medium, 2 - none
enum Mode fr_cur_mode
mode index of current frame
float prediction_error[4]
quantified prediction errors {20log10(^gamma_gc)} for previous four subframes
float tilt_coef
{beta_1} related to the voicing of the previous subframe
double isp_sub4_past[LP_ORDER]
ISP vector for the 4th subframe of the previous frame.
float pitch_vector[AMRWB_SFR_SIZE]
adaptive codebook (pitch) vector for current subframe
float lp_coef[4][LP_ORDER]
Linear Prediction Coefficients from ISP vector.
float demph_mem[1]
previous value in the de-emphasis filter
float fixed_gain[2]
quantified fixed gains for the current and previous subframes
float isf_cur[LP_ORDER]
working ISF vector from current frame
float samples_up[UPS_MEM_SIZE+AMRWB_SFR_SIZE]
low-band samples and memory processed for upsampling
uint16_t isp_id[7]
index of ISP subvectors
uint16_t vad
voice activity detection flag
AMRWBSubFrame subframe[4]
data for subframes
uint16_t adap
adaptive codebook index
uint16_t vq_gain
VQ adaptive and innovative gains.
uint16_t hb_gain
high-band energy index (mode 23k85 only)
uint16_t ltp
ltp-filtering flag
uint16_t pul_il[4]
LSBs part of codebook index.
uint16_t pul_ih[4]
MSBs part of codebook index (high modes only)
main external API structure.
enum AVSampleFormat sample_fmt
audio sample format
int sample_rate
samples per second
int channels
number of audio channels
uint64_t channel_layout
Audio channel layout.
const char * name
Name of the codec implementation.
This structure describes decoded (raw) audio or video data.
int nb_samples
number of audio samples (per channel) described by this frame
uint8_t * data[AV_NUM_DATA_POINTERS]
pointer to the picture/channel planes.
Context structure for the Lagged Fibonacci PRNG.
This structure stores compressed data.
#define avpriv_request_sample(...)
static float mean(const float *input, int size)