Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(1034)

Issues Search
    My Issues | Starred     Open | Closed | All

Unified Diff: src/opus_encoder.c

Issue 28553003: Updating Opus to a pre-release of 1.1 (Closed) Base URL: svn://svn.chromium.org/chrome/trunk/deps/third_party/opus
Patch Set: Removing failing file Created 7 years, 2 months ago
Use n/p to move between diff chunks; N/P to move between comments. Draft comments are only viewable by you.
Jump to:
View side-by-side diff with in-line comments
Download patch
« no previous file with comments | « src/opus_demo.c ('k') | src/opus_multistream.c » ('j') | no next file with comments »
Expand Comments ('e') | Collapse Comments ('c') | Show Comments Hide Comments ('s')
Index: src/opus_encoder.c
diff --git a/src/opus_encoder.c b/src/opus_encoder.c
index aae31256ca579a7642b474adc97516d37080b85c..f6f0ce5c3ad162e7a0fa67ed28d98b85b8549f1d 100644
--- a/src/opus_encoder.c
+++ b/src/opus_encoder.c
@@ -40,7 +40,9 @@
#include "arch.h"
#include "opus_private.h"
#include "os_support.h"
-
+#include "cpu_support.h"
+#include "analysis.h"
+#include "mathops.h"
#include "tuning_parameters.h"
#ifdef FIXED_POINT
#include "fixed/structs_FIX.h"
@@ -50,6 +52,12 @@
#define MAX_ENCODER_BUFFER 480
+typedef struct {
+ opus_val32 XX, XY, YY;
+ opus_val16 smoothed_width;
+ opus_val16 max_follower;
+} StereoWidthState;
+
struct OpusEncoder {
int celt_enc_offset;
int silk_enc_offset;
@@ -66,14 +74,18 @@ struct OpusEncoder {
opus_int32 Fs;
int use_vbr;
int vbr_constraint;
+ int variable_duration;
opus_int32 bitrate_bps;
opus_int32 user_bitrate_bps;
+ int lsb_depth;
int encoder_buffer;
+ int lfe;
#define OPUS_ENCODER_RESET_START stream_channels
int stream_channels;
opus_int16 hybrid_stereo_width_Q14;
opus_int32 variable_HP_smth2_Q15;
+ opus_val16 prev_HB_gain;
opus_val32 hp_mem[4];
int mode;
int prev_mode;
@@ -83,9 +95,16 @@ struct OpusEncoder {
int silk_bw_switch;
/* Sampling rate (at the API level) */
int first;
+ opus_val16 * energy_masking;
+ StereoWidthState width_mem;
opus_val16 delay_buffer[MAX_ENCODER_BUFFER*2];
-
+#ifndef DISABLE_FLOAT_API
+ TonalityAnalysisState analysis;
+ int detected_bandwidth;
+ int analysis_offset;
+#endif
opus_uint32 rangeFinal;
+ int arch;
};
/* Transition tables for the voice and music. First column is the
@@ -94,8 +113,8 @@ struct OpusEncoder {
static const opus_int32 mono_voice_bandwidth_thresholds[8] = {
11000, 1000, /* NB<->MB */
14000, 1000, /* MB<->WB */
- 21000, 2000, /* WB<->SWB */
- 29000, 2000, /* SWB<->FB */
+ 17000, 1000, /* WB<->SWB */
+ 20000, 1000, /* SWB<->FB */
};
static const opus_int32 mono_music_bandwidth_thresholds[8] = {
14000, 1000, /* MB not allowed */
@@ -116,14 +135,14 @@ static const opus_int32 stereo_music_bandwidth_thresholds[8] = {
48000, 2000, /* SWB<->FB */
};
/* Threshold bit-rates for switching between mono and stereo */
-static const opus_int32 stereo_voice_threshold = 26000;
-static const opus_int32 stereo_music_threshold = 36000;
+static const opus_int32 stereo_voice_threshold = 31000;
+static const opus_int32 stereo_music_threshold = 31000;
/* Threshold bit-rate for switching between SILK/hybrid and CELT-only */
static const opus_int32 mode_thresholds[2][2] = {
/* voice */ /* music */
- { 48000, 24000}, /* mono */
- { 48000, 24000}, /* stereo */
+ { 64000, 20000}, /* mono */
+ { 36000, 20000}, /* stereo */
};
int opus_encoder_get_size(int channels)
@@ -167,6 +186,8 @@ int opus_encoder_init(OpusEncoder* st, opus_int32 Fs, int channels, int applicat
st->Fs = Fs;
+ st->arch = opus_select_arch();
+
ret = silk_InitEncoder( silk_enc, &st->silk_mode );
if(ret)return OPUS_INTERNAL_ERROR;
@@ -180,7 +201,7 @@ int opus_encoder_init(OpusEncoder* st, opus_int32 Fs, int channels, int applicat
st->silk_mode.payloadSize_ms = 20;
st->silk_mode.bitRate = 25000;
st->silk_mode.packetLossPercentage = 0;
- st->silk_mode.complexity = 10;
+ st->silk_mode.complexity = 9;
st->silk_mode.useInBandFEC = 0;
st->silk_mode.useDTX = 0;
st->silk_mode.useCBR = 0;
@@ -191,7 +212,7 @@ int opus_encoder_init(OpusEncoder* st, opus_int32 Fs, int channels, int applicat
if(err!=OPUS_OK)return OPUS_INTERNAL_ERROR;
celt_encoder_ctl(celt_enc, CELT_SET_SIGNALLING(0));
- celt_encoder_ctl(celt_enc, OPUS_SET_COMPLEXITY(10));
+ celt_encoder_ctl(celt_enc, OPUS_SET_COMPLEXITY(st->silk_mode.complexity));
st->use_vbr = 1;
/* Makes constrained VBR the default (safer for real-time use) */
@@ -206,12 +227,15 @@ int opus_encoder_init(OpusEncoder* st, opus_int32 Fs, int channels, int applicat
st->user_forced_mode = OPUS_AUTO;
st->voice_ratio = -1;
st->encoder_buffer = st->Fs/100;
+ st->lsb_depth = 24;
+ st->variable_duration = OPUS_FRAMESIZE_ARG;
/* Delay compensation of 4 ms (2.5 ms for SILK's extra look-ahead
+ 1.5 ms for SILK resamplers and stereo prediction) */
st->delay_compensation = st->Fs/250;
st->hybrid_stereo_width_Q14 = 1 << 14;
+ st->prev_HB_gain = Q15ONE;
st->variable_HP_smth2_Q15 = silk_LSHIFT( silk_lin2log( VARIABLE_HP_MIN_CUTOFF_HZ ), 8 );
st->first = 1;
st->mode = MODE_HYBRID;
@@ -322,7 +346,7 @@ static void silk_biquad_float(
S[ 0 ] = S[1] - vout*A[0] + B[1]*inval;
- S[ 1 ] = - vout*A[1] + B[2]*inval;
+ S[ 1 ] = - vout*A[1] + B[2]*inval + VERY_SMALL;
/* Scale back to Q0 and saturate */
out[ k*stride ] = vout;
@@ -365,6 +389,56 @@ static void hp_cutoff(const opus_val16 *in, opus_int32 cutoff_Hz, opus_val16 *ou
#endif
}
+#ifdef FIXED_POINT
+static void dc_reject(const opus_val16 *in, opus_int32 cutoff_Hz, opus_val16 *out, opus_val32 *hp_mem, int len, int channels, opus_int32 Fs)
+{
+ int c, i;
+ int shift;
+
+ /* Approximates -round(log2(4.*cutoff_Hz/Fs)) */
+ shift=celt_ilog2(Fs/(cutoff_Hz*3));
+ for (c=0;c<channels;c++)
+ {
+ for (i=0;i<len;i++)
+ {
+ opus_val32 x, tmp, y;
+ x = SHL32(EXTEND32(in[channels*i+c]), 15);
+ /* First stage */
+ tmp = x-hp_mem[2*c];
+ hp_mem[2*c] = hp_mem[2*c] + PSHR32(x - hp_mem[2*c], shift);
+ /* Second stage */
+ y = tmp - hp_mem[2*c+1];
+ hp_mem[2*c+1] = hp_mem[2*c+1] + PSHR32(tmp - hp_mem[2*c+1], shift);
+ out[channels*i+c] = EXTRACT16(SATURATE(PSHR32(y, 15), 32767));
+ }
+ }
+}
+
+#else
+static void dc_reject(const opus_val16 *in, opus_int32 cutoff_Hz, opus_val16 *out, opus_val32 *hp_mem, int len, int channels, opus_int32 Fs)
+{
+ int c, i;
+ float coef;
+
+ coef = 4.0f*cutoff_Hz/Fs;
+ for (c=0;c<channels;c++)
+ {
+ for (i=0;i<len;i++)
+ {
+ opus_val32 x, tmp, y;
+ x = in[channels*i+c];
+ /* First stage */
+ tmp = x-hp_mem[2*c];
+ hp_mem[2*c] = hp_mem[2*c] + coef*(x - hp_mem[2*c]) + VERY_SMALL;
+ /* Second stage */
+ y = tmp - hp_mem[2*c+1];
+ hp_mem[2*c+1] = hp_mem[2*c+1] + coef*(tmp - hp_mem[2*c+1]) + VERY_SMALL;
+ out[channels*i+c] = y;
+ }
+ }
+}
+#endif
+
static void stereo_fade(const opus_val16 *in, opus_val16 *out, opus_val16 g1, opus_val16 g2,
int overlap48, int frame_size, int channels, const opus_val16 *window, opus_int32 Fs)
{
@@ -397,6 +471,45 @@ static void stereo_fade(const opus_val16 *in, opus_val16 *out, opus_val16 g1, op
}
}
+static void gain_fade(const opus_val16 *in, opus_val16 *out, opus_val16 g1, opus_val16 g2,
+ int overlap48, int frame_size, int channels, const opus_val16 *window, opus_int32 Fs)
+{
+ int i;
+ int inc;
+ int overlap;
+ int c;
+ inc = 48000/Fs;
+ overlap=overlap48/inc;
+ if (channels==1)
+ {
+ for (i=0;i<overlap;i++)
+ {
+ opus_val16 g, w;
+ w = MULT16_16_Q15(window[i*inc], window[i*inc]);
+ g = SHR32(MAC16_16(MULT16_16(w,g2),
+ Q15ONE-w, g1), 15);
+ out[i] = MULT16_16_Q15(g, in[i]);
+ }
+ } else {
+ for (i=0;i<overlap;i++)
+ {
+ opus_val16 g, w;
+ w = MULT16_16_Q15(window[i*inc], window[i*inc]);
+ g = SHR32(MAC16_16(MULT16_16(w,g2),
+ Q15ONE-w, g1), 15);
+ out[i*2] = MULT16_16_Q15(g, in[i*2]);
+ out[i*2+1] = MULT16_16_Q15(g, in[i*2+1]);
+ }
+ }
+ c=0;do {
+ for (i=overlap;i<frame_size;i++)
+ {
+ out[i*channels+c] = MULT16_16_Q15(g2, in[i*channels+c]);
+ }
+ }
+ while (++c<channels);
+}
+
OpusEncoder *opus_encoder_create(opus_int32 Fs, int channels, int application, int *error)
{
int ret;
@@ -438,15 +551,412 @@ static opus_int32 user_bitrate_to_bitrate(OpusEncoder *st, int frame_size, int m
return st->user_bitrate_bps;
}
+#ifndef DISABLE_FLOAT_API
+/* Don't use more than 60 ms for the frame size analysis */
+#define MAX_DYNAMIC_FRAMESIZE 24
+/* Estimates how much the bitrate will be boosted based on the sub-frame energy */
+static float transient_boost(const float *E, const float *E_1, int LM, int maxM)
+{
+ int i;
+ int M;
+ float sumE=0, sumE_1=0;
+ float metric;
+
+ M = IMIN(maxM, (1<<LM)+1);
+ for (i=0;i<M;i++)
+ {
+ sumE += E[i];
+ sumE_1 += E_1[i];
+ }
+ metric = sumE*sumE_1/(M*M);
+ /*if (LM==3)
+ printf("%f\n", metric);*/
+ /*return metric>10 ? 1 : 0;*/
+ /*return MAX16(0,1-exp(-.25*(metric-2.)));*/
+ return MIN16(1,(float)sqrt(MAX16(0,.05f*(metric-2))));
+}
+
+/* Viterbi decoding trying to find the best frame size combination using look-ahead
+
+ State numbering:
+ 0: unused
+ 1: 2.5 ms
+ 2: 5 ms (#1)
+ 3: 5 ms (#2)
+ 4: 10 ms (#1)
+ 5: 10 ms (#2)
+ 6: 10 ms (#3)
+ 7: 10 ms (#4)
+ 8: 20 ms (#1)
+ 9: 20 ms (#2)
+ 10: 20 ms (#3)
+ 11: 20 ms (#4)
+ 12: 20 ms (#5)
+ 13: 20 ms (#6)
+ 14: 20 ms (#7)
+ 15: 20 ms (#8)
+*/
+static int transient_viterbi(const float *E, const float *E_1, int N, int frame_cost, int rate)
+{
+ int i;
+ float cost[MAX_DYNAMIC_FRAMESIZE][16];
+ int states[MAX_DYNAMIC_FRAMESIZE][16];
+ float best_cost;
+ int best_state;
+ float factor;
+ /* Take into account that we damp VBR in the 32 kb/s to 64 kb/s range. */
+ if (rate<80)
+ factor=0;
+ else if (rate>160)
+ factor=1;
+ else
+ factor = (rate-80.f)/80.f;
+ /* Makes variable framesize less aggressive at lower bitrates, but I can't
+ find any valid theoretical justification for this (other than it seems
+ to help) */
+ for (i=0;i<16;i++)
+ {
+ /* Impossible state */
+ states[0][i] = -1;
+ cost[0][i] = 1e10;
+ }
+ for (i=0;i<4;i++)
+ {
+ cost[0][1<<i] = (frame_cost + rate*(1<<i))*(1+factor*transient_boost(E, E_1, i, N+1));
+ states[0][1<<i] = i;
+ }
+ for (i=1;i<N;i++)
+ {
+ int j;
+
+ /* Follow continuations */
+ for (j=2;j<16;j++)
+ {
+ cost[i][j] = cost[i-1][j-1];
+ states[i][j] = j-1;
+ }
+
+ /* New frames */
+ for(j=0;j<4;j++)
+ {
+ int k;
+ float min_cost;
+ float curr_cost;
+ states[i][1<<j] = 1;
+ min_cost = cost[i-1][1];
+ for(k=1;k<4;k++)
+ {
+ float tmp = cost[i-1][(1<<(k+1))-1];
+ if (tmp < min_cost)
+ {
+ states[i][1<<j] = (1<<(k+1))-1;
+ min_cost = tmp;
+ }
+ }
+ curr_cost = (frame_cost + rate*(1<<j))*(1+factor*transient_boost(E+i, E_1+i, j, N-i+1));
+ cost[i][1<<j] = min_cost;
+ /* If part of the frame is outside the analysis window, only count part of the cost */
+ if (N-i < (1<<j))
+ cost[i][1<<j] += curr_cost*(float)(N-i)/(1<<j);
+ else
+ cost[i][1<<j] += curr_cost;
+ }
+ }
+
+ best_state=1;
+ best_cost = cost[N-1][1];
+ /* Find best end state (doesn't force a frame to end at N-1) */
+ for (i=2;i<16;i++)
+ {
+ if (cost[N-1][i]<best_cost)
+ {
+ best_cost = cost[N-1][i];
+ best_state = i;
+ }
+ }
+
+ /* Follow transitions back */
+ for (i=N-1;i>=0;i--)
+ {
+ /*printf("%d ", best_state);*/
+ best_state = states[i][best_state];
+ }
+ /*printf("%d\n", best_state);*/
+ return best_state;
+}
+
+int optimize_framesize(const opus_val16 *x, int len, int C, opus_int32 Fs,
+ int bitrate, opus_val16 tonality, opus_val32 *mem, int buffering,
+ downmix_func downmix)
+{
+ int N;
+ int i;
+ float e[MAX_DYNAMIC_FRAMESIZE+4];
+ float e_1[MAX_DYNAMIC_FRAMESIZE+3];
+ float memx;
+ int bestLM=0;
+ int subframe;
+ int pos;
+ VARDECL(opus_val32, sub);
+
+ subframe = Fs/400;
+ ALLOC(sub, subframe, opus_val32);
+ e[0]=mem[0];
+ e_1[0]=1.f/(EPSILON+mem[0]);
+ if (buffering)
+ {
+ /* Consider the CELT delay when not in restricted-lowdelay */
+ /* We assume the buffering is between 2.5 and 5 ms */
+ int offset = 2*subframe - buffering;
+ celt_assert(offset>=0 && offset <= subframe);
+ x += C*offset;
+ len -= offset;
+ e[1]=mem[1];
+ e_1[1]=1.f/(EPSILON+mem[1]);
+ e[2]=mem[2];
+ e_1[2]=1.f/(EPSILON+mem[2]);
+ pos = 3;
+ } else {
+ pos=1;
+ }
+ N=IMIN(len/subframe, MAX_DYNAMIC_FRAMESIZE);
+ memx = x[0];
+ for (i=0;i<N;i++)
+ {
+ float tmp;
+ float tmpx;
+ int j;
+ tmp=EPSILON;
+
+ downmix(x, sub, subframe, i*subframe, 0, -2, C);
+ if (i==0)
+ memx = sub[0];
+ for (j=0;j<subframe;j++)
+ {
+ tmpx = sub[j];
+ tmp += (tmpx-memx)*(tmpx-memx);
+ memx = tmpx;
+ }
+ e[i+pos] = tmp;
+ e_1[i+pos] = 1.f/tmp;
+ }
+ /* Hack to get 20 ms working with APPLICATION_AUDIO
+ The real problem is that the corresponding memory needs to use 1.5 ms
+ from this frame and 1 ms from the next frame */
+ e[i+pos] = e[i+pos-1];
+ if (buffering)
+ N=IMIN(MAX_DYNAMIC_FRAMESIZE, N+2);
+ bestLM = transient_viterbi(e, e_1, N, (int)((1.f+.5f*tonality)*(60*C+40)), bitrate/400);
+ mem[0] = e[1<<bestLM];
+ if (buffering)
+ {
+ mem[1] = e[(1<<bestLM)+1];
+ mem[2] = e[(1<<bestLM)+2];
+ }
+ return bestLM;
+}
+
+#endif
+
+#ifndef DISABLE_FLOAT_API
#ifdef FIXED_POINT
-#define opus_encode_native opus_encode
-opus_int32 opus_encode(OpusEncoder *st, const opus_val16 *pcm, int frame_size,
- unsigned char *data, opus_int32 out_data_bytes)
+#define PCM2VAL(x) FLOAT2INT16(x)
#else
-#define opus_encode_native opus_encode_float
-opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_size,
- unsigned char *data, opus_int32 out_data_bytes)
+#define PCM2VAL(x) SCALEIN(x)
+#endif
+void downmix_float(const void *_x, opus_val32 *sub, int subframe, int offset, int c1, int c2, int C)
+{
+ const float *x;
+ opus_val32 scale;
+ int j;
+ x = (const float *)_x;
+ for (j=0;j<subframe;j++)
+ sub[j] = PCM2VAL(x[(j+offset)*C+c1]);
+ if (c2>-1)
+ {
+ for (j=0;j<subframe;j++)
+ sub[j] += PCM2VAL(x[(j+offset)*C+c2]);
+ } else if (c2==-2)
+ {
+ int c;
+ for (c=1;c<C;c++)
+ {
+ for (j=0;j<subframe;j++)
+ sub[j] += PCM2VAL(x[(j+offset)*C+c]);
+ }
+ }
+#ifdef FIXED_POINT
+ scale = (1<<SIG_SHIFT);
+#else
+ scale = 1.f;
+#endif
+ if (C==-2)
+ scale /= C;
+ else
+ scale /= 2;
+ for (j=0;j<subframe;j++)
+ sub[j] *= scale;
+}
+#endif
+
+void downmix_int(const void *_x, opus_val32 *sub, int subframe, int offset, int c1, int c2, int C)
+{
+ const opus_int16 *x;
+ opus_val32 scale;
+ int j;
+ x = (const opus_int16 *)_x;
+ for (j=0;j<subframe;j++)
+ sub[j] = x[(j+offset)*C+c1];
+ if (c2>-1)
+ {
+ for (j=0;j<subframe;j++)
+ sub[j] += x[(j+offset)*C+c2];
+ } else if (c2==-2)
+ {
+ int c;
+ for (c=1;c<C;c++)
+ {
+ for (j=0;j<subframe;j++)
+ sub[j] += x[(j+offset)*C+c];
+ }
+ }
+#ifdef FIXED_POINT
+ scale = (1<<SIG_SHIFT);
+#else
+ scale = 1.f/32768;
+#endif
+ if (C==-2)
+ scale /= C;
+ else
+ scale /= 2;
+ for (j=0;j<subframe;j++)
+ sub[j] *= scale;
+}
+
+opus_int32 frame_size_select(opus_int32 frame_size, int variable_duration, opus_int32 Fs)
+{
+ int new_size;
+ if (frame_size<Fs/400)
+ return -1;
+ if (variable_duration == OPUS_FRAMESIZE_ARG)
+ new_size = frame_size;
+ else if (variable_duration == OPUS_FRAMESIZE_VARIABLE)
+ new_size = Fs/50;
+ else if (variable_duration >= OPUS_FRAMESIZE_2_5_MS && variable_duration <= OPUS_FRAMESIZE_60_MS)
+ new_size = IMIN(3*Fs/50, (Fs/400)<<(variable_duration-OPUS_FRAMESIZE_2_5_MS));
+ else
+ return -1;
+ if (new_size>frame_size)
+ return -1;
+ if (400*new_size!=Fs && 200*new_size!=Fs && 100*new_size!=Fs &&
+ 50*new_size!=Fs && 25*new_size!=Fs && 50*new_size!=3*Fs)
+ return -1;
+ return new_size;
+}
+
+opus_int32 compute_frame_size(const void *analysis_pcm, int frame_size,
+ int variable_duration, int C, opus_int32 Fs, int bitrate_bps,
+ int delay_compensation, downmix_func downmix, opus_val32 *subframe_mem)
+{
+#ifndef DISABLE_FLOAT_API
+ if (variable_duration == OPUS_FRAMESIZE_VARIABLE && frame_size >= Fs/200)
+ {
+ int LM = 3;
+ LM = optimize_framesize(analysis_pcm, frame_size, C, Fs, bitrate_bps,
+ 0, subframe_mem, delay_compensation, downmix);
+ while ((Fs/400<<LM)>frame_size)
+ LM--;
+ frame_size = (Fs/400<<LM);
+ } else
#endif
+ {
+ frame_size = frame_size_select(frame_size, variable_duration, Fs);
+ }
+ if (frame_size<0)
+ return -1;
+ return frame_size;
+}
+
+opus_val16 compute_stereo_width(const opus_val16 *pcm, int frame_size, opus_int32 Fs, StereoWidthState *mem)
+{
+ opus_val16 corr;
+ opus_val16 ldiff;
+ opus_val16 width;
+ opus_val32 xx, xy, yy;
+ opus_val16 sqrt_xx, sqrt_yy;
+ opus_val16 qrrt_xx, qrrt_yy;
+ int frame_rate;
+ int i;
+ opus_val16 short_alpha;
+
+ frame_rate = Fs/frame_size;
+ short_alpha = Q15ONE - 25*Q15ONE/IMAX(50,frame_rate);
+ xx=xy=yy=0;
+ for (i=0;i<frame_size;i+=4)
+ {
+ opus_val32 pxx=0;
+ opus_val32 pxy=0;
+ opus_val32 pyy=0;
+ opus_val16 x, y;
+ x = pcm[2*i];
+ y = pcm[2*i+1];
+ pxx = SHR32(MULT16_16(x,x),2);
+ pxy = SHR32(MULT16_16(x,y),2);
+ pyy = SHR32(MULT16_16(y,y),2);
+ x = pcm[2*i+2];
+ y = pcm[2*i+3];
+ pxx += SHR32(MULT16_16(x,x),2);
+ pxy += SHR32(MULT16_16(x,y),2);
+ pyy += SHR32(MULT16_16(y,y),2);
+ x = pcm[2*i+4];
+ y = pcm[2*i+5];
+ pxx += SHR32(MULT16_16(x,x),2);
+ pxy += SHR32(MULT16_16(x,y),2);
+ pyy += SHR32(MULT16_16(y,y),2);
+ x = pcm[2*i+6];
+ y = pcm[2*i+7];
+ pxx += SHR32(MULT16_16(x,x),2);
+ pxy += SHR32(MULT16_16(x,y),2);
+ pyy += SHR32(MULT16_16(y,y),2);
+
+ xx += SHR32(pxx, 10);
+ xy += SHR32(pxy, 10);
+ yy += SHR32(pyy, 10);
+ }
+ mem->XX += MULT16_32_Q15(short_alpha, xx-mem->XX);
+ mem->XY += MULT16_32_Q15(short_alpha, xy-mem->XY);
+ mem->YY += MULT16_32_Q15(short_alpha, yy-mem->YY);
+ mem->XX = MAX32(0, mem->XX);
+ mem->XY = MAX32(0, mem->XY);
+ mem->YY = MAX32(0, mem->YY);
+ if (MAX32(mem->XX, mem->YY)>QCONST16(8e-4f, 18))
+ {
+ sqrt_xx = celt_sqrt(mem->XX);
+ sqrt_yy = celt_sqrt(mem->YY);
+ qrrt_xx = celt_sqrt(sqrt_xx);
+ qrrt_yy = celt_sqrt(sqrt_yy);
+ /* Inter-channel correlation */
+ mem->XY = MIN32(mem->XY, sqrt_xx*sqrt_yy);
+ corr = SHR32(frac_div32(mem->XY,EPSILON+MULT16_16(sqrt_xx,sqrt_yy)),16);
+ /* Approximate loudness difference */
+ ldiff = Q15ONE*ABS16(qrrt_xx-qrrt_yy)/(EPSILON+qrrt_xx+qrrt_yy);
+ width = MULT16_16_Q15(celt_sqrt(QCONST32(1.f,30)-MULT16_16(corr,corr)), ldiff);
+ /* Smoothing over one second */
+ mem->smoothed_width += (width-mem->smoothed_width)/frame_rate;
+ /* Peak follower */
+ mem->max_follower = MAX16(mem->max_follower-QCONST16(.02f,15)/frame_rate, mem->smoothed_width);
+ } else {
+ width = 0;
+ corr=Q15ONE;
+ ldiff=0;
+ }
+ /*printf("%f %f %f %f %f ", corr/(float)Q15ONE, ldiff/(float)Q15ONE, width/(float)Q15ONE, mem->smoothed_width/(float)Q15ONE, mem->max_follower/(float)Q15ONE);*/
+ return EXTRACT16(MIN32(Q15ONE,20*mem->max_follower));
+}
+
+opus_int32 opus_encode_native(OpusEncoder *st, const opus_val16 *pcm, int frame_size,
+ unsigned char *data, opus_int32 out_data_bytes, int lsb_depth,
+ const void *analysis_pcm, opus_int32 analysis_size, int c1, int c2, int analysis_channels, downmix_func downmix)
{
void *silk_enc;
CELTEncoder *celt_enc;
@@ -471,7 +981,14 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s
int frame_rate;
opus_int32 max_rate; /* Max bitrate we're allowed to use */
int curr_bandwidth;
+ opus_val16 HB_gain;
opus_int32 max_data_bytes; /* Max number of bytes we're allowed to use */
+ int total_buffer;
+ opus_val16 stereo_width;
+ const CELTMode *celt_mode;
+ AnalysisInfo analysis_info;
+ int analysis_read_pos_bak=-1;
+ int analysis_read_subframe_bak=-1;
VARDECL(opus_val16, tmp_prefill);
ALLOC_STACK;
@@ -479,25 +996,70 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s
max_data_bytes = IMIN(1276, out_data_bytes);
st->rangeFinal = 0;
- if (400*frame_size != st->Fs && 200*frame_size != st->Fs && 100*frame_size != st->Fs &&
+ if ((!st->variable_duration && 400*frame_size != st->Fs && 200*frame_size != st->Fs && 100*frame_size != st->Fs &&
50*frame_size != st->Fs && 25*frame_size != st->Fs && 50*frame_size != 3*st->Fs)
- {
- RESTORE_STACK;
- return OPUS_BAD_ARG;
- }
- if (max_data_bytes<=0)
+ || (400*frame_size < st->Fs)
+ || max_data_bytes<=0
+ )
{
RESTORE_STACK;
return OPUS_BAD_ARG;
}
silk_enc = (char*)st+st->silk_enc_offset;
celt_enc = (CELTEncoder*)((char*)st+st->celt_enc_offset);
-
if (st->application == OPUS_APPLICATION_RESTRICTED_LOWDELAY)
delay_compensation = 0;
else
delay_compensation = st->delay_compensation;
+ lsb_depth = IMIN(lsb_depth, st->lsb_depth);
+
+ analysis_info.valid = 0;
+ celt_encoder_ctl(celt_enc, CELT_GET_MODE(&celt_mode));
+#ifndef DISABLE_FLOAT_API
+#ifdef FIXED_POINT
+ if (st->silk_mode.complexity >= 10 && st->Fs==48000)
+#else
+ if (st->silk_mode.complexity >= 7 && st->Fs==48000)
+#endif
+ {
+ analysis_read_pos_bak = st->analysis.read_pos;
+ analysis_read_subframe_bak = st->analysis.read_subframe;
+ run_analysis(&st->analysis, celt_mode, analysis_pcm, analysis_size, frame_size,
+ c1, c2, analysis_channels, st->Fs,
+ lsb_depth, downmix, &analysis_info);
+ }
+#endif
+
+ st->voice_ratio = -1;
+
+#ifndef DISABLE_FLOAT_API
+ st->detected_bandwidth = 0;
+ if (analysis_info.valid)
+ {
+ int analysis_bandwidth;
+ if (st->signal_type == OPUS_AUTO)
+ st->voice_ratio = (int)floor(.5+100*(1-analysis_info.music_prob));
+
+ analysis_bandwidth = analysis_info.bandwidth;
+ if (analysis_bandwidth<=12)
+ st->detected_bandwidth = OPUS_BANDWIDTH_NARROWBAND;
+ else if (analysis_bandwidth<=14)
+ st->detected_bandwidth = OPUS_BANDWIDTH_MEDIUMBAND;
+ else if (analysis_bandwidth<=16)
+ st->detected_bandwidth = OPUS_BANDWIDTH_WIDEBAND;
+ else if (analysis_bandwidth<=18)
+ st->detected_bandwidth = OPUS_BANDWIDTH_SUPERWIDEBAND;
+ else
+ st->detected_bandwidth = OPUS_BANDWIDTH_FULLBAND;
+ }
+#endif
+
+ if (st->channels==2 && st->force_channels!=1)
+ stereo_width = compute_stereo_width(pcm, frame_size, st->Fs, &st->width_mem);
+ else
+ stereo_width = 0;
+ total_buffer = delay_compensation;
st->bitrate_bps = user_bitrate_to_bitrate(st, frame_size, max_data_bytes);
frame_rate = st->Fs/frame_size;
@@ -540,8 +1102,12 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s
else if (st->signal_type == OPUS_SIGNAL_MUSIC)
voice_est = 0;
else if (st->voice_ratio >= 0)
+ {
voice_est = st->voice_ratio*327>>8;
- else if (st->application == OPUS_APPLICATION_VOIP)
+ /* For AUDIO, never be more than 90% confident of having speech */
+ if (st->application == OPUS_APPLICATION_AUDIO)
+ voice_est = IMIN(voice_est, 115);
+ } else if (st->application == OPUS_APPLICATION_VOIP)
voice_est = 115;
else
voice_est = 48;
@@ -561,9 +1127,9 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s
opus_int32 stereo_threshold;
stereo_threshold = stereo_music_threshold + ((voice_est*voice_est*(stereo_voice_threshold-stereo_music_threshold))>>14);
if (st->stream_channels == 2)
- stereo_threshold -= 4000;
+ stereo_threshold -= 1000;
else
- stereo_threshold += 4000;
+ stereo_threshold += 1000;
st->stream_channels = (equiv_rate > stereo_threshold) ? 2 : 1;
} else {
st->stream_channels = st->channels;
@@ -592,15 +1158,21 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s
st->mode = MODE_SILK_ONLY;
}
#else
- int chan;
opus_int32 mode_voice, mode_music;
opus_int32 threshold;
- chan = (st->channels==2) && st->force_channels!=1;
- mode_voice = mode_thresholds[chan][0];
- mode_music = mode_thresholds[chan][1];
+ /* Interpolate based on stereo width */
+ mode_voice = (opus_int32)(MULT16_32_Q15(Q15ONE-stereo_width,mode_thresholds[0][0])
+ + MULT16_32_Q15(stereo_width,mode_thresholds[1][0]));
+ mode_music = (opus_int32)(MULT16_32_Q15(Q15ONE-stereo_width,mode_thresholds[1][1])
+ + MULT16_32_Q15(stereo_width,mode_thresholds[1][1]));
+ /* Interpolate based on speech/music probability */
threshold = mode_music + ((voice_est*voice_est*(mode_voice-mode_music))>>14);
+ /* Bias towards SILK for VoIP because of some useful features */
+ if (st->application == OPUS_APPLICATION_VOIP)
+ threshold += 8000;
+ /*printf("%f %d\n", stereo_width/(float)Q15ONE, threshold);*/
/* Hysteresis */
if (st->prev_mode == MODE_CELT_ONLY)
threshold -= 4000;
@@ -658,6 +1230,7 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s
redundancy = 1;
celt_to_silk = 1;
st->silk_bw_switch = 0;
+ prefill=1;
}
if (redundancy)
@@ -750,6 +1323,31 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s
st->bandwidth = OPUS_BANDWIDTH_MEDIUMBAND;
if (st->Fs <= 8000 && st->bandwidth > OPUS_BANDWIDTH_NARROWBAND)
st->bandwidth = OPUS_BANDWIDTH_NARROWBAND;
+#ifndef FIXED_POINT
+ /* Use detected bandwidth to reduce the encoded bandwidth. */
+ if (st->detected_bandwidth && st->user_bandwidth == OPUS_AUTO)
+ {
+ int min_detected_bandwidth;
+ /* Makes bandwidth detection more conservative just in case the detector
+ gets it wrong when we could have coded a high bandwidth transparently.
+ When operating in SILK/hybrid mode, we don't go below wideband to avoid
+ more complicated switches that require redundancy. */
+ if (st->bitrate_bps <= 18000*st->stream_channels && st->mode == MODE_CELT_ONLY)
+ min_detected_bandwidth = OPUS_BANDWIDTH_NARROWBAND;
+ else if (st->bitrate_bps <= 24000*st->stream_channels && st->mode == MODE_CELT_ONLY)
+ min_detected_bandwidth = OPUS_BANDWIDTH_MEDIUMBAND;
+ else if (st->bitrate_bps <= 30000*st->stream_channels)
+ min_detected_bandwidth = OPUS_BANDWIDTH_WIDEBAND;
+ else if (st->bitrate_bps <= 44000*st->stream_channels)
+ min_detected_bandwidth = OPUS_BANDWIDTH_SUPERWIDEBAND;
+ else
+ min_detected_bandwidth = OPUS_BANDWIDTH_FULLBAND;
+
+ st->detected_bandwidth = IMAX(st->detected_bandwidth, min_detected_bandwidth);
+ st->bandwidth = IMIN(st->bandwidth, st->detected_bandwidth);
+ }
+#endif
+ celt_encoder_ctl(celt_enc, OPUS_SET_LSB_DEPTH(lsb_depth));
/* If max_data_bytes represents less than 8 kb/s, switch to CELT-only mode */
if (max_data_bytes < (frame_rate > 50 ? 12000 : 8000)*frame_size / (st->Fs * 8))
@@ -758,6 +1356,11 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s
/* CELT mode doesn't support mediumband, use wideband instead */
if (st->mode == MODE_CELT_ONLY && st->bandwidth == OPUS_BANDWIDTH_MEDIUMBAND)
st->bandwidth = OPUS_BANDWIDTH_WIDEBAND;
+ if (st->lfe)
+ {
+ st->bandwidth = OPUS_BANDWIDTH_NARROWBAND;
+ st->mode = MODE_CELT_ONLY;
+ }
/* Can't support higher than wideband for >20 ms frames */
if (frame_size > st->Fs/50 && (st->mode == MODE_CELT_ONLY || st->bandwidth > OPUS_BANDWIDTH_WIDEBAND))
@@ -765,16 +1368,22 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s
VARDECL(unsigned char, tmp_data);
int nb_frames;
int bak_mode, bak_bandwidth, bak_channels, bak_to_mono;
- OpusRepacketizer rp;
+ VARDECL(OpusRepacketizer, rp);
opus_int32 bytes_per_frame;
+ if (analysis_read_pos_bak!= -1)
+ {
+ st->analysis.read_pos = analysis_read_pos_bak;
+ st->analysis.read_subframe = analysis_read_subframe_bak;
+ }
nb_frames = frame_size > st->Fs/25 ? 3 : 2;
bytes_per_frame = IMIN(1276,(out_data_bytes-3)/nb_frames);
ALLOC(tmp_data, nb_frames*bytes_per_frame, unsigned char);
- opus_repacketizer_init(&rp);
+ ALLOC(rp, 1, OpusRepacketizer);
+ opus_repacketizer_init(rp);
bak_mode = st->user_forced_mode;
bak_bandwidth = st->user_bandwidth;
@@ -796,20 +1405,22 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s
/* When switching from SILK/Hybrid to CELT, only ask for a switch at the last frame */
if (to_celt && i==nb_frames-1)
st->user_forced_mode = MODE_CELT_ONLY;
- tmp_len = opus_encode_native(st, pcm+i*(st->channels*st->Fs/50), st->Fs/50, tmp_data+i*bytes_per_frame, bytes_per_frame);
+ tmp_len = opus_encode_native(st, pcm+i*(st->channels*st->Fs/50), st->Fs/50,
+ tmp_data+i*bytes_per_frame, bytes_per_frame, lsb_depth,
+ NULL, 0, c1, c2, analysis_channels, downmix);
if (tmp_len<0)
{
RESTORE_STACK;
return OPUS_INTERNAL_ERROR;
}
- ret = opus_repacketizer_cat(&rp, tmp_data+i*bytes_per_frame, tmp_len);
+ ret = opus_repacketizer_cat(rp, tmp_data+i*bytes_per_frame, tmp_len);
if (ret<0)
{
RESTORE_STACK;
return OPUS_INTERNAL_ERROR;
}
}
- ret = opus_repacketizer_out(&rp, data, out_data_bytes);
+ ret = opus_repacketizer_out(rp, data, out_data_bytes);
if (ret<0)
{
RESTORE_STACK;
@@ -822,7 +1433,6 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s
RESTORE_STACK;
return ret;
}
-
curr_bandwidth = st->bandwidth;
/* Chooses the appropriate mode for speech
@@ -839,9 +1449,9 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s
ec_enc_init(&enc, data, max_data_bytes-1);
- ALLOC(pcm_buf, (delay_compensation+frame_size)*st->channels, opus_val16);
- for (i=0;i<delay_compensation*st->channels;i++)
- pcm_buf[i] = st->delay_buffer[(st->encoder_buffer-delay_compensation)*st->channels+i];
+ ALLOC(pcm_buf, (total_buffer+frame_size)*st->channels, opus_val16);
+ for (i=0;i<total_buffer*st->channels;i++)
+ pcm_buf[i] = st->delay_buffer[(st->encoder_buffer-total_buffer)*st->channels+i];
if (st->mode == MODE_CELT_ONLY)
hp_freq_smth1 = silk_LSHIFT( silk_lin2log( VARIABLE_HP_MIN_CUTOFF_HZ ), 8 );
@@ -856,46 +1466,89 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s
if (st->application == OPUS_APPLICATION_VOIP)
{
- hp_cutoff(pcm, cutoff_Hz, &pcm_buf[delay_compensation*st->channels], st->hp_mem, frame_size, st->channels, st->Fs);
+ hp_cutoff(pcm, cutoff_Hz, &pcm_buf[total_buffer*st->channels], st->hp_mem, frame_size, st->channels, st->Fs);
} else {
- for (i=0;i<frame_size*st->channels;i++)
- pcm_buf[delay_compensation*st->channels + i] = pcm[i];
+ dc_reject(pcm, 3, &pcm_buf[total_buffer*st->channels], st->hp_mem, frame_size, st->channels, st->Fs);
}
+
+
/* SILK processing */
+ HB_gain = Q15ONE;
if (st->mode != MODE_CELT_ONLY)
{
+ opus_int32 total_bitRate, celt_rate;
#ifdef FIXED_POINT
const opus_int16 *pcm_silk;
#else
VARDECL(opus_int16, pcm_silk);
ALLOC(pcm_silk, st->channels*frame_size, opus_int16);
#endif
- st->silk_mode.bitRate = 8*bytes_target*frame_rate;
+
+ /* Distribute bits between SILK and CELT */
+ total_bitRate = 8 * bytes_target * frame_rate;
if( st->mode == MODE_HYBRID ) {
- st->silk_mode.bitRate /= st->stream_channels;
+ int HB_gain_ref;
+ /* Base rate for SILK */
+ st->silk_mode.bitRate = st->stream_channels * ( 5000 + 1000 * ( st->Fs == 100 * frame_size ) );
if( curr_bandwidth == OPUS_BANDWIDTH_SUPERWIDEBAND ) {
- if( st->Fs == 100 * frame_size ) {
- /* 24 kHz, 10 ms */
- st->silk_mode.bitRate = ( ( st->silk_mode.bitRate + 2000 + st->use_vbr * 1000 ) * 2 ) / 3;
- } else {
- /* 24 kHz, 20 ms */
- st->silk_mode.bitRate = ( ( st->silk_mode.bitRate + 1000 + st->use_vbr * 1000 ) * 2 ) / 3;
- }
- } else {
- if( st->Fs == 100 * frame_size ) {
- /* 48 kHz, 10 ms */
- st->silk_mode.bitRate = ( st->silk_mode.bitRate + 8000 + st->use_vbr * 3000 ) / 2;
- } else {
- /* 48 kHz, 20 ms */
- st->silk_mode.bitRate = ( st->silk_mode.bitRate + 9000 + st->use_vbr * 1000 ) / 2;
- }
+ /* SILK gets 2/3 of the remaining bits */
+ st->silk_mode.bitRate += ( total_bitRate - st->silk_mode.bitRate ) * 2 / 3;
+ } else { /* FULLBAND */
+ /* SILK gets 3/5 of the remaining bits */
+ st->silk_mode.bitRate += ( total_bitRate - st->silk_mode.bitRate ) * 3 / 5;
}
- st->silk_mode.bitRate *= st->stream_channels;
- /* don't let SILK use more than 80% */
- if( st->silk_mode.bitRate > ( st->bitrate_bps - 8*st->Fs/frame_size ) * 4/5 ) {
- st->silk_mode.bitRate = ( st->bitrate_bps - 8*st->Fs/frame_size ) * 4/5;
+ /* Don't let SILK use more than 80% */
+ if( st->silk_mode.bitRate > total_bitRate * 4/5 ) {
+ st->silk_mode.bitRate = total_bitRate * 4/5;
}
+ /* Increasingly attenuate high band when it gets allocated fewer bits */
+ celt_rate = total_bitRate - st->silk_mode.bitRate;
+ HB_gain_ref = (curr_bandwidth == OPUS_BANDWIDTH_SUPERWIDEBAND) ? 3000 : 3600;
+ HB_gain = SHL32((opus_val32)celt_rate, 9) / SHR32((opus_val32)celt_rate + st->stream_channels * HB_gain_ref, 6);
+ HB_gain = HB_gain < Q15ONE*6/7 ? HB_gain + Q15ONE/7 : Q15ONE;
+ } else {
+ /* SILK gets all bits */
+ st->silk_mode.bitRate = total_bitRate;
+ }
+
+ /* Surround masking for SILK */
+ if (st->energy_masking && st->use_vbr && !st->lfe)
+ {
+ opus_val32 mask_sum=0;
+ opus_val16 masking_depth;
+ opus_int32 rate_offset;
+ int c;
+ int end = 17;
+ opus_int16 srate = 16000;
+ if (st->bandwidth == OPUS_BANDWIDTH_NARROWBAND)
+ {
+ end = 13;
+ srate = 8000;
+ } else if (st->bandwidth == OPUS_BANDWIDTH_MEDIUMBAND)
+ {
+ end = 15;
+ srate = 12000;
+ }
+ for (c=0;c<st->channels;c++)
+ {
+ for(i=0;i<end;i++)
+ {
+ opus_val16 mask;
+ mask = MAX16(MIN16(st->energy_masking[21*c+i],
+ QCONST16(.25f, DB_SHIFT)), -QCONST16(2.0f, DB_SHIFT));
+ if (mask > 0)
+ mask = HALF16(mask);
+ mask_sum += mask;
+ }
+ }
+ /* Conservative rate reduction, we cut the masking in half */
+ masking_depth = HALF16(mask_sum / end*st->channels);
+ rate_offset = PSHR32(MULT16_16(srate, masking_depth), DB_SHIFT);
+ rate_offset = MAX32(rate_offset, -2*st->silk_mode.bitRate/3);
+ rate_offset += QCONST16(.4f, DB_SHIFT);
+ st->silk_mode.bitRate += rate_offset;
+ bytes_target += rate_offset * frame_size / (8 * st->Fs);
}
st->silk_mode.payloadSize_ms = 1000 * frame_size / st->Fs;
@@ -955,6 +1608,18 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s
if (prefill)
{
opus_int32 zero=0;
+ int prefill_offset;
+ /* Use a smooth onset for the SILK prefill to avoid the encoder trying to encode
+ a discontinuity. The exact location is what we need to avoid leaving any "gap"
+ in the audio when mixing with the redundant CELT frame. Here we can afford to
+ overwrite st->delay_buffer because the only thing that uses it before it gets
+ rewritten is tmp_prefill[] and even then only the part after the ramp really
+ gets used (rather than sent to the encoder and discarded) */
+ prefill_offset = st->channels*(st->encoder_buffer-st->delay_compensation-st->Fs/400);
+ gain_fade(st->delay_buffer+prefill_offset, st->delay_buffer+prefill_offset,
+ 0, Q15ONE, celt_mode->overlap, st->Fs/400, st->channels, celt_mode->window, st->Fs);
+ for(i=0;i<prefill_offset;i++)
+ st->delay_buffer[i]=0;
#ifdef FIXED_POINT
pcm_silk = st->delay_buffer;
#else
@@ -965,10 +1630,10 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s
}
#ifdef FIXED_POINT
- pcm_silk = pcm_buf+delay_compensation*st->channels;
+ pcm_silk = pcm_buf+total_buffer*st->channels;
#else
for (i=0;i<frame_size*st->channels;i++)
- pcm_silk[i] = FLOAT2INT16(pcm_buf[delay_compensation*st->channels + i]);
+ pcm_silk[i] = FLOAT2INT16(pcm_buf[total_buffer*st->channels + i]);
#endif
ret = silk_Encode( silk_enc, &st->silk_mode, pcm_silk, frame_size, &enc, &nBytes, 0 );
if( ret ) {
@@ -1053,9 +1718,18 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s
} else {
if (st->use_vbr)
{
+ opus_int32 bonus=0;
+#ifndef DISABLE_FLOAT_API
+ if (st->variable_duration==OPUS_FRAMESIZE_VARIABLE && frame_size != st->Fs/50)
+ {
+ bonus = (60*st->stream_channels+40)*(st->Fs/frame_size-50);
+ if (analysis_info.valid)
+ bonus = (opus_int32)(bonus*(1.f+.5f*analysis_info.tonality));
+ }
+#endif
celt_encoder_ctl(celt_enc, OPUS_SET_VBR(1));
celt_encoder_ctl(celt_enc, OPUS_SET_VBR_CONSTRAINT(st->vbr_constraint));
- celt_encoder_ctl(celt_enc, OPUS_SET_BITRATE(st->bitrate_bps));
+ celt_encoder_ctl(celt_enc, OPUS_SET_BITRATE(st->bitrate_bps+bonus));
nb_compr_bytes = max_data_bytes-1-redundancy_bytes;
} else {
nb_compr_bytes = bytes_target;
@@ -1070,24 +1744,27 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s
if (st->mode != MODE_SILK_ONLY && st->mode != st->prev_mode && st->prev_mode > 0)
{
for (i=0;i<st->channels*st->Fs/400;i++)
- tmp_prefill[i] = st->delay_buffer[(st->encoder_buffer-st->delay_compensation-st->Fs/400)*st->channels + i];
+ tmp_prefill[i] = st->delay_buffer[(st->encoder_buffer-total_buffer-st->Fs/400)*st->channels + i];
}
- for (i=0;i<st->channels*(st->encoder_buffer-(frame_size+delay_compensation));i++)
+ for (i=0;i<st->channels*(st->encoder_buffer-(frame_size+total_buffer));i++)
st->delay_buffer[i] = st->delay_buffer[i+st->channels*frame_size];
for (;i<st->encoder_buffer*st->channels;i++)
- st->delay_buffer[i] = pcm_buf[(frame_size+delay_compensation-st->encoder_buffer)*st->channels+i];
-
+ st->delay_buffer[i] = pcm_buf[(frame_size+total_buffer-st->encoder_buffer)*st->channels+i];
+ /* gain_fade() and stereo_fade() need to be after the buffer copying
+ because we don't want any of this to affect the SILK part */
+ if( st->prev_HB_gain < Q15ONE || HB_gain < Q15ONE ) {
+ gain_fade(pcm_buf, pcm_buf,
+ st->prev_HB_gain, HB_gain, celt_mode->overlap, frame_size, st->channels, celt_mode->window, st->Fs);
+ }
+ st->prev_HB_gain = HB_gain;
if (st->mode != MODE_HYBRID || st->stream_channels==1)
- st->silk_mode.stereoWidth_Q14 = 1<<14;
- if( st->channels == 2 ) {
+ st->silk_mode.stereoWidth_Q14 = IMIN((1<<14),IMAX(0,st->bitrate_bps-32000));
+ if( !st->energy_masking && st->channels == 2 ) {
/* Apply stereo width reduction (at low bitrates) */
if( st->hybrid_stereo_width_Q14 < (1 << 14) || st->silk_mode.stereoWidth_Q14 < (1 << 14) ) {
opus_val16 g1, g2;
- const CELTMode *celt_mode;
-
- celt_encoder_ctl(celt_enc, CELT_GET_MODE(&celt_mode));
g1 = st->hybrid_stereo_width_Q14;
g2 = (opus_val16)(st->silk_mode.stereoWidth_Q14);
#ifdef FIXED_POINT
@@ -1144,6 +1821,10 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s
ec_enc_shrink(&enc, nb_compr_bytes);
}
+#ifndef DISABLE_FLOAT_API
+ if (redundancy || st->mode != MODE_SILK_ONLY)
+ celt_encoder_ctl(celt_enc, CELT_SET_ANALYSIS(&analysis_info));
+#endif
/* 5 ms redundant frame for CELT->SILK */
if (redundancy && celt_to_silk)
@@ -1268,45 +1949,89 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s
#ifdef FIXED_POINT
#ifndef DISABLE_FLOAT_API
-opus_int32 opus_encode_float(OpusEncoder *st, const float *pcm, int frame_size,
+opus_int32 opus_encode_float(OpusEncoder *st, const float *pcm, int analysis_frame_size,
unsigned char *data, opus_int32 max_data_bytes)
{
int i, ret;
+ int frame_size;
+ int delay_compensation;
VARDECL(opus_int16, in);
ALLOC_STACK;
- if(frame_size<0)
- {
- RESTORE_STACK;
- return OPUS_BAD_ARG;
- }
+ if (st->application == OPUS_APPLICATION_RESTRICTED_LOWDELAY)
+ delay_compensation = 0;
+ else
+ delay_compensation = st->delay_compensation;
+ frame_size = compute_frame_size(pcm, analysis_frame_size,
+ st->variable_duration, st->channels, st->Fs, st->bitrate_bps,
+ delay_compensation, downmix_float, st->analysis.subframe_mem);
ALLOC(in, frame_size*st->channels, opus_int16);
for (i=0;i<frame_size*st->channels;i++)
in[i] = FLOAT2INT16(pcm[i]);
- ret = opus_encode(st, in, frame_size, data, max_data_bytes);
+ ret = opus_encode_native(st, in, frame_size, data, max_data_bytes, 16, pcm, analysis_frame_size, 0, -2, st->channels, downmix_float);
RESTORE_STACK;
return ret;
}
#endif
+opus_int32 opus_encode(OpusEncoder *st, const opus_int16 *pcm, int analysis_frame_size,
+ unsigned char *data, opus_int32 out_data_bytes)
+{
+ int frame_size;
+ int delay_compensation;
+ if (st->application == OPUS_APPLICATION_RESTRICTED_LOWDELAY)
+ delay_compensation = 0;
+ else
+ delay_compensation = st->delay_compensation;
+ frame_size = compute_frame_size(pcm, analysis_frame_size,
+ st->variable_duration, st->channels, st->Fs, st->bitrate_bps,
+ delay_compensation, downmix_float, st->analysis.subframe_mem);
+ return opus_encode_native(st, pcm, frame_size, data, out_data_bytes, 16, pcm, analysis_frame_size, 0, -2, st->channels, downmix_int);
+}
+
#else
-opus_int32 opus_encode(OpusEncoder *st, const opus_int16 *pcm, int frame_size,
+opus_int32 opus_encode(OpusEncoder *st, const opus_int16 *pcm, int analysis_frame_size,
unsigned char *data, opus_int32 max_data_bytes)
{
int i, ret;
+ int frame_size;
+ int delay_compensation;
VARDECL(float, in);
ALLOC_STACK;
+ if (st->application == OPUS_APPLICATION_RESTRICTED_LOWDELAY)
+ delay_compensation = 0;
+ else
+ delay_compensation = st->delay_compensation;
+ frame_size = compute_frame_size(pcm, analysis_frame_size,
+ st->variable_duration, st->channels, st->Fs, st->bitrate_bps,
+ delay_compensation, downmix_float, st->analysis.subframe_mem);
+
ALLOC(in, frame_size*st->channels, float);
for (i=0;i<frame_size*st->channels;i++)
in[i] = (1.0f/32768)*pcm[i];
- ret = opus_encode_float(st, in, frame_size, data, max_data_bytes);
+ ret = opus_encode_native(st, in, frame_size, data, max_data_bytes, 16, pcm, analysis_frame_size, 0, -2, st->channels, downmix_int);
RESTORE_STACK;
return ret;
}
+opus_int32 opus_encode_float(OpusEncoder *st, const float *pcm, int analysis_frame_size,
+ unsigned char *data, opus_int32 out_data_bytes)
+{
+ int frame_size;
+ int delay_compensation;
+ if (st->application == OPUS_APPLICATION_RESTRICTED_LOWDELAY)
+ delay_compensation = 0;
+ else
+ delay_compensation = st->delay_compensation;
+ frame_size = compute_frame_size(pcm, analysis_frame_size,
+ st->variable_duration, st->channels, st->Fs, st->bitrate_bps,
+ delay_compensation, downmix_float, st->analysis.subframe_mem);
+ return opus_encode_native(st, pcm, frame_size, data, out_data_bytes, 24,
+ pcm, analysis_frame_size, 0, -2, st->channels, downmix_float);
+}
#endif
@@ -1339,6 +2064,10 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...)
case OPUS_GET_APPLICATION_REQUEST:
{
opus_int32 *value = va_arg(ap, opus_int32*);
+ if (!value)
+ {
+ goto bad_arg;
+ }
*value = st->application;
}
break;
@@ -1360,6 +2089,10 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...)
case OPUS_GET_BITRATE_REQUEST:
{
opus_int32 *value = va_arg(ap, opus_int32*);
+ if (!value)
+ {
+ goto bad_arg;
+ }
*value = user_bitrate_to_bitrate(st, st->prev_framesize, 1276);
}
break;
@@ -1367,21 +2100,29 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...)
{
opus_int32 value = va_arg(ap, opus_int32);
if((value<1 || value>st->channels) && value != OPUS_AUTO)
- return OPUS_BAD_ARG;
+ {
+ goto bad_arg;
+ }
st->force_channels = value;
}
break;
case OPUS_GET_FORCE_CHANNELS_REQUEST:
{
opus_int32 *value = va_arg(ap, opus_int32*);
+ if (!value)
+ {
+ goto bad_arg;
+ }
*value = st->force_channels;
}
break;
case OPUS_SET_MAX_BANDWIDTH_REQUEST:
{
opus_int32 value = va_arg(ap, opus_int32);
- if (value < OPUS_BANDWIDTH_NARROWBAND || value > OPUS_BANDWIDTH_FULLBAND)
- return OPUS_BAD_ARG;
+ if (value < OPUS_BANDWIDTH_NARROWBAND || value > OPUS_BANDWIDTH_FULLBAND)
+ {
+ goto bad_arg;
+ }
st->max_bandwidth = value;
if (st->max_bandwidth == OPUS_BANDWIDTH_NARROWBAND) {
st->silk_mode.maxInternalSampleRate = 8000;
@@ -1395,6 +2136,10 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...)
case OPUS_GET_MAX_BANDWIDTH_REQUEST:
{
opus_int32 *value = va_arg(ap, opus_int32*);
+ if (!value)
+ {
+ goto bad_arg;
+ }
*value = st->max_bandwidth;
}
break;
@@ -1402,7 +2147,9 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...)
{
opus_int32 value = va_arg(ap, opus_int32);
if ((value < OPUS_BANDWIDTH_NARROWBAND || value > OPUS_BANDWIDTH_FULLBAND) && value != OPUS_AUTO)
- return OPUS_BAD_ARG;
+ {
+ goto bad_arg;
+ }
st->user_bandwidth = value;
if (st->user_bandwidth == OPUS_BANDWIDTH_NARROWBAND) {
st->silk_mode.maxInternalSampleRate = 8000;
@@ -1416,6 +2163,10 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...)
case OPUS_GET_BANDWIDTH_REQUEST:
{
opus_int32 *value = va_arg(ap, opus_int32*);
+ if (!value)
+ {
+ goto bad_arg;
+ }
*value = st->bandwidth;
}
break;
@@ -1423,13 +2174,19 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...)
{
opus_int32 value = va_arg(ap, opus_int32);
if(value<0 || value>1)
- return OPUS_BAD_ARG;
+ {
+ goto bad_arg;
+ }
st->silk_mode.useDTX = value;
}
break;
case OPUS_GET_DTX_REQUEST:
{
opus_int32 *value = va_arg(ap, opus_int32*);
+ if (!value)
+ {
+ goto bad_arg;
+ }
*value = st->silk_mode.useDTX;
}
break;
@@ -1437,7 +2194,9 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...)
{
opus_int32 value = va_arg(ap, opus_int32);
if(value<0 || value>10)
- return OPUS_BAD_ARG;
+ {
+ goto bad_arg;
+ }
st->silk_mode.complexity = value;
celt_encoder_ctl(celt_enc, OPUS_SET_COMPLEXITY(value));
}
@@ -1445,6 +2204,10 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...)
case OPUS_GET_COMPLEXITY_REQUEST:
{
opus_int32 *value = va_arg(ap, opus_int32*);
+ if (!value)
+ {
+ goto bad_arg;
+ }
*value = st->silk_mode.complexity;
}
break;
@@ -1452,13 +2215,19 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...)
{
opus_int32 value = va_arg(ap, opus_int32);
if(value<0 || value>1)
- return OPUS_BAD_ARG;
+ {
+ goto bad_arg;
+ }
st->silk_mode.useInBandFEC = value;
}
break;
case OPUS_GET_INBAND_FEC_REQUEST:
{
opus_int32 *value = va_arg(ap, opus_int32*);
+ if (!value)
+ {
+ goto bad_arg;
+ }
*value = st->silk_mode.useInBandFEC;
}
break;
@@ -1466,7 +2235,9 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...)
{
opus_int32 value = va_arg(ap, opus_int32);
if (value < 0 || value > 100)
- return OPUS_BAD_ARG;
+ {
+ goto bad_arg;
+ }
st->silk_mode.packetLossPercentage = value;
celt_encoder_ctl(celt_enc, OPUS_SET_PACKET_LOSS_PERC(value));
}
@@ -1474,6 +2245,10 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...)
case OPUS_GET_PACKET_LOSS_PERC_REQUEST:
{
opus_int32 *value = va_arg(ap, opus_int32*);
+ if (!value)
+ {
+ goto bad_arg;
+ }
*value = st->silk_mode.packetLossPercentage;
}
break;
@@ -1481,7 +2256,9 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...)
{
opus_int32 value = va_arg(ap, opus_int32);
if(value<0 || value>1)
- return OPUS_BAD_ARG;
+ {
+ goto bad_arg;
+ }
st->use_vbr = value;
st->silk_mode.useCBR = 1-value;
}
@@ -1489,20 +2266,30 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...)
case OPUS_GET_VBR_REQUEST:
{
opus_int32 *value = va_arg(ap, opus_int32*);
+ if (!value)
+ {
+ goto bad_arg;
+ }
*value = st->use_vbr;
}
break;
case OPUS_SET_VOICE_RATIO_REQUEST:
{
opus_int32 value = va_arg(ap, opus_int32);
- if (value>100 || value<-1)
- goto bad_arg;
+ if (value<-1 || value>100)
+ {
+ goto bad_arg;
+ }
st->voice_ratio = value;
}
break;
case OPUS_GET_VOICE_RATIO_REQUEST:
{
opus_int32 *value = va_arg(ap, opus_int32*);
+ if (!value)
+ {
+ goto bad_arg;
+ }
*value = st->voice_ratio;
}
break;
@@ -1510,13 +2297,19 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...)
{
opus_int32 value = va_arg(ap, opus_int32);
if(value<0 || value>1)
- return OPUS_BAD_ARG;
+ {
+ goto bad_arg;
+ }
st->vbr_constraint = value;
}
break;
case OPUS_GET_VBR_CONSTRAINT_REQUEST:
{
opus_int32 *value = va_arg(ap, opus_int32*);
+ if (!value)
+ {
+ goto bad_arg;
+ }
*value = st->vbr_constraint;
}
break;
@@ -1524,19 +2317,29 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...)
{
opus_int32 value = va_arg(ap, opus_int32);
if(value!=OPUS_AUTO && value!=OPUS_SIGNAL_VOICE && value!=OPUS_SIGNAL_MUSIC)
- return OPUS_BAD_ARG;
+ {
+ goto bad_arg;
+ }
st->signal_type = value;
}
break;
case OPUS_GET_SIGNAL_REQUEST:
{
opus_int32 *value = va_arg(ap, opus_int32*);
+ if (!value)
+ {
+ goto bad_arg;
+ }
*value = st->signal_type;
}
break;
case OPUS_GET_LOOKAHEAD_REQUEST:
{
opus_int32 *value = va_arg(ap, opus_int32*);
+ if (!value)
+ {
+ goto bad_arg;
+ }
*value = st->Fs/400;
if (st->application != OPUS_APPLICATION_RESTRICTED_LOWDELAY)
*value += st->delay_compensation;
@@ -1545,10 +2348,9 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...)
case OPUS_GET_SAMPLE_RATE_REQUEST:
{
opus_int32 *value = va_arg(ap, opus_int32*);
- if (value==NULL)
+ if (!value)
{
- ret = OPUS_BAD_ARG;
- break;
+ goto bad_arg;
}
*value = st->Fs;
}
@@ -1556,19 +2358,55 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...)
case OPUS_GET_FINAL_RANGE_REQUEST:
{
opus_uint32 *value = va_arg(ap, opus_uint32*);
+ if (!value)
+ {
+ goto bad_arg;
+ }
*value = st->rangeFinal;
}
break;
case OPUS_SET_LSB_DEPTH_REQUEST:
{
opus_int32 value = va_arg(ap, opus_int32);
- ret = celt_encoder_ctl(celt_enc, OPUS_SET_LSB_DEPTH(value));
+ if (value<8 || value>24)
+ {
+ goto bad_arg;
+ }
+ st->lsb_depth=value;
}
break;
case OPUS_GET_LSB_DEPTH_REQUEST:
{
opus_int32 *value = va_arg(ap, opus_int32*);
- celt_encoder_ctl(celt_enc, OPUS_GET_LSB_DEPTH(value));
+ if (!value)
+ {
+ goto bad_arg;
+ }
+ *value = st->lsb_depth;
+ }
+ break;
+ case OPUS_SET_EXPERT_FRAME_DURATION_REQUEST:
+ {
+ opus_int32 value = va_arg(ap, opus_int32);
+ if (value != OPUS_FRAMESIZE_ARG && value != OPUS_FRAMESIZE_2_5_MS &&
+ value != OPUS_FRAMESIZE_5_MS && value != OPUS_FRAMESIZE_10_MS &&
+ value != OPUS_FRAMESIZE_20_MS && value != OPUS_FRAMESIZE_40_MS &&
+ value != OPUS_FRAMESIZE_60_MS && value != OPUS_FRAMESIZE_VARIABLE)
+ {
+ goto bad_arg;
+ }
+ st->variable_duration = value;
+ celt_encoder_ctl(celt_enc, OPUS_SET_EXPERT_FRAME_DURATION(value));
+ }
+ break;
+ case OPUS_GET_EXPERT_FRAME_DURATION_REQUEST:
+ {
+ opus_int32 *value = va_arg(ap, opus_int32*);
+ if (!value)
+ {
+ goto bad_arg;
+ }
+ *value = st->variable_duration;
}
break;
case OPUS_RESET_STATE:
@@ -1585,6 +2423,7 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...)
silk_InitEncoder( silk_enc, &dummy );
st->stream_channels = st->channels;
st->hybrid_stereo_width_Q14 = 1 << 14;
+ st->prev_HB_gain = Q15ONE;
st->first = 1;
st->mode = MODE_HYBRID;
st->bandwidth = OPUS_BANDWIDTH_FULLBAND;
@@ -1595,10 +2434,37 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...)
{
opus_int32 value = va_arg(ap, opus_int32);
if ((value < MODE_SILK_ONLY || value > MODE_CELT_ONLY) && value != OPUS_AUTO)
+ {
goto bad_arg;
+ }
st->user_forced_mode = value;
}
break;
+ case OPUS_SET_LFE_REQUEST:
+ {
+ opus_int32 value = va_arg(ap, opus_int32);
+ st->lfe = value;
+ ret = celt_encoder_ctl(celt_enc, OPUS_SET_LFE(value));
+ }
+ break;
+ case OPUS_SET_ENERGY_MASK_REQUEST:
+ {
+ opus_val16 *value = va_arg(ap, opus_val16*);
+ st->energy_masking = value;
+ ret = celt_encoder_ctl(celt_enc, OPUS_SET_ENERGY_MASK(value));
+ }
+ break;
+
+ case CELT_GET_MODE_REQUEST:
+ {
+ const CELTMode ** value = va_arg(ap, const CELTMode**);
+ if (!value)
+ {
+ goto bad_arg;
+ }
+ ret = celt_encoder_ctl(celt_enc, CELT_GET_MODE(value));
+ }
+ break;
default:
/* fprintf(stderr, "unknown opus_encoder_ctl() request: %d", request);*/
ret = OPUS_UNIMPLEMENTED;
« no previous file with comments | « src/opus_demo.c ('k') | src/opus_multistream.c » ('j') | no next file with comments »

Powered by Google App Engine
This is Rietveld 408576698