[xiph-commits] r14137 - in trunk/speex: include/speex libspeex
jm at svn.xiph.org
jm at svn.xiph.org
Wed Nov 14 03:59:22 PST 2007
Author: jm
Date: 2007-11-14 03:59:22 -0800 (Wed, 14 Nov 2007)
New Revision: 14137
Modified:
trunk/speex/include/speex/speex_stereo.h
trunk/speex/libspeex/stereo.c
trunk/speex/libspeex/vq.c
trunk/speex/libspeex/vq.h
Log:
fixed-point: Patch by Thom Johansen to convert stereo decoding to fixed-point,
plus, my own ugly API hack to actually maintain compatibility.
Modified: trunk/speex/include/speex/speex_stereo.h
===================================================================
--- trunk/speex/include/speex/speex_stereo.h 2007-11-13 15:42:33 UTC (rev 14136)
+++ trunk/speex/include/speex/speex_stereo.h 2007-11-14 11:59:22 UTC (rev 14137)
@@ -46,7 +46,7 @@
extern "C" {
#endif
-/** State used for decoding (intensity) stereo information */
+/** If you access any of these fields directly, I'll personally come and bite you */
typedef struct SpeexStereoState {
float balance; /**< Left/right balance info */
float e_ratio; /**< Ratio of energies: E(left+right)/[E(left)+E(right)] */
@@ -56,9 +56,18 @@
float reserved2; /**< Reserved for future use */
} SpeexStereoState;
-/** Initialization value for a stereo state */
+/** Deprecated. Use speex_stereo_state_init() instead. */
#define SPEEX_STEREO_STATE_INIT {1,.5,1,1,0,0}
+/** Initialise/create a stereo stereo state */
+SpeexStereoState *speex_stereo_state_init();
+
+/** Reset/re-initialise an already allocated stereo state */
+void speex_stereo_state_reset(SpeexStereoState *stereo);
+
+/** Destroy a stereo stereo state */
+void speex_stereo_state_destroy(SpeexStereoState *stereo);
+
/** Transforms a stereo frame into a mono frame and stores intensity stereo info in 'bits' */
void speex_encode_stereo(float *data, int frame_size, SpeexBits *bits);
Modified: trunk/speex/libspeex/stereo.c
===================================================================
--- trunk/speex/libspeex/stereo.c 2007-11-13 15:42:33 UTC (rev 14136)
+++ trunk/speex/libspeex/stereo.c 2007-11-14 11:59:22 UTC (rev 14137)
@@ -35,12 +35,71 @@
#include <speex/speex_stereo.h>
#include <speex/speex_callbacks.h>
+#include "math_approx.h"
#include "vq.h"
#include <math.h>
+#include "os_support.h"
+typedef struct RealSpeexStereoState {
+ spx_word32_t balance; /**< Left/right balance info */
+ spx_word32_t e_ratio; /**< Ratio of energies: E(left+right)/[E(left)+E(right)] */
+ spx_word32_t smooth_left; /**< Smoothed left channel gain */
+ spx_word32_t smooth_right; /**< Smoothed right channel gain */
+ spx_uint32_t reserved1; /**< Reserved for future use */
+ spx_int32_t reserved2; /**< Reserved for future use */
+} RealSpeexStereoState;
+
+
/*float e_ratio_quant[4] = {1, 1.26, 1.587, 2};*/
+#ifndef FIXED_POINT
static const float e_ratio_quant[4] = {.25f, .315f, .397f, .5f};
+static const float e_ratio_quant_bounds[3] = {0.2825f, 0.356f, 0.4485f};
+#else
+static const spx_word16_t e_ratio_quant[4] = {8192, 10332, 13009, 16384};
+static const spx_word16_t e_ratio_quant_bounds[3] = {9257, 11665, 14696};
+#endif
+/* This is an ugly compatibility hack that properly resets the stereo state
+ In case it it compiled in fixed-point, but initialised with the deprecated
+ floating point static initialiser */
+#ifdef FIXED_POINT
+#define COMPATIBILITY_HACK(s) do {if ((s)->reserved1 != 0xdeadbeef) speex_stereo_state_reset((SpeexStereoState*)s); } while (0);
+#else
+#define COMPATIBILITY_HACK(s)
+#endif
+
+SpeexStereoState *speex_stereo_state_init()
+{
+ SpeexStereoState *stereo = speex_alloc(sizeof(SpeexStereoState));
+ speex_stereo_state_reset(stereo);
+ return stereo;
+}
+
+void speex_stereo_state_reset(SpeexStereoState *_stereo)
+{
+ RealSpeexStereoState *stereo = (RealSpeexStereoState*)_stereo;
+#ifdef FIXED_POINT
+ stereo->balance = 65536;
+ stereo->e_ratio = 16384;
+ stereo->smooth_left = 16384;
+ stereo->smooth_right = 16384;
+ stereo->reserved1 = 0xdeadbeef;
+ stereo->reserved2 = 0;
+#else
+ stereo->balance = 1.0f;
+ stereo->e_ratio = .5f;
+ stereo->smooth_left = 1.f;
+ stereo->smooth_right = 1.f;
+ stereo->reserved1 = 0;
+ stereo->reserved2 = 0;
+#endif
+}
+
+void speex_stereo_state_destroy(SpeexStereoState *stereo)
+{
+ speex_free(stereo);
+}
+
void speex_encode_stereo(float *data, int frame_size, SpeexBits *bits)
{
int i, tmp;
@@ -73,8 +132,8 @@
speex_bits_pack(bits, (int)balance, 5);
- /*Quantize energy ratio*/
- tmp=vq_index(&e_ratio, e_ratio_quant, 1, 4);
+ /* FIXME: this is a hack */
+ tmp=scal_quant(e_ratio*Q15_ONE, e_ratio_quant_bounds, 3);
speex_bits_pack(bits, tmp, 2);
}
@@ -110,81 +169,81 @@
speex_bits_pack(bits, (int)balance, 5);
- /*Quantize energy ratio*/
- tmp=vq_index(&e_ratio, e_ratio_quant, 1, 4);
+ /* FIXME: this is a hack */
+ tmp=scal_quant(e_ratio*Q15_ONE, e_ratio_quant_bounds, 3);
speex_bits_pack(bits, tmp, 2);
}
-void speex_decode_stereo(float *data, int frame_size, SpeexStereoState *stereo)
+void speex_decode_stereo(float *data, int frame_size, SpeexStereoState *_stereo)
{
- float balance, e_ratio;
int i;
- float e_tot=0, e_left, e_right, e_sum;
-
+ spx_word32_t balance;
+ spx_word16_t e_left, e_right, e_ratio;
+ RealSpeexStereoState *stereo = (RealSpeexStereoState*)_stereo;
+
+ COMPATIBILITY_HACK(stereo);
+
balance=stereo->balance;
e_ratio=stereo->e_ratio;
- for (i=frame_size-1;i>=0;i--)
- {
- e_tot += ((float)data[i])*data[i];
- }
- e_sum=e_tot/e_ratio;
- e_left = e_sum*balance / (1+balance);
- e_right = e_sum-e_left;
+
+ /* These two are Q14, with max value just below 2. */
+ e_right = DIV32(QCONST32(1., 22), spx_sqrt(MULT16_32_Q15(e_ratio, ADD32(QCONST32(1., 16), balance))));
+ e_left = SHR32(MULT16_16(spx_sqrt(balance), e_right), 8);
- e_left = sqrt(e_left/(e_tot+.01));
- e_right = sqrt(e_right/(e_tot+.01));
-
for (i=frame_size-1;i>=0;i--)
{
- float ftmp=data[i];
- stereo->smooth_left = .98*stereo->smooth_left + .02*e_left;
- stereo->smooth_right = .98*stereo->smooth_right + .02*e_right;
- data[2*i] = stereo->smooth_left*ftmp;
- data[2*i+1] = stereo->smooth_right*ftmp;
+ spx_word16_t tmp=data[i];
+ stereo->smooth_left = EXTRACT16(PSHR32(MAC16_16(MULT16_16(stereo->smooth_left, QCONST16(0.98, 15)), e_left, QCONST16(0.02, 15)), 15));
+ stereo->smooth_right = EXTRACT16(PSHR32(MAC16_16(MULT16_16(stereo->smooth_right, QCONST16(0.98, 15)), e_right, QCONST16(0.02, 15)), 15));
+ data[2*i] = (float)MULT16_16_P14(stereo->smooth_left, tmp);
+ data[2*i+1] = (float)MULT16_16_P14(stereo->smooth_right, tmp);
}
}
-void speex_decode_stereo_int(spx_int16_t *data, int frame_size, SpeexStereoState *stereo)
+void speex_decode_stereo_int(spx_int16_t *data, int frame_size, SpeexStereoState *_stereo)
{
- float balance, e_ratio;
int i;
- float e_tot=0, e_left, e_right, e_sum;
+ spx_word32_t balance;
+ spx_word16_t e_left, e_right, e_ratio;
+ RealSpeexStereoState *stereo = (RealSpeexStereoState*)_stereo;
+ COMPATIBILITY_HACK(stereo);
+
balance=stereo->balance;
e_ratio=stereo->e_ratio;
- for (i=frame_size-1;i>=0;i--)
- {
- e_tot += ((float)data[i])*data[i];
- }
- e_sum=e_tot/e_ratio;
- e_left = e_sum*balance / (1+balance);
- e_right = e_sum-e_left;
+
+ /* These two are Q14, with max value just below 2. */
+ e_right = DIV32(QCONST32(1., 22), spx_sqrt(MULT16_32_Q15(e_ratio, ADD32(QCONST32(1., 16), balance))));
+ e_left = SHR32(MULT16_16(spx_sqrt(balance), e_right), 8);
- e_left = sqrt(e_left/(e_tot+.01));
- e_right = sqrt(e_right/(e_tot+.01));
-
for (i=frame_size-1;i>=0;i--)
{
- float ftmp=data[i];
- stereo->smooth_left = .98*stereo->smooth_left + .02*e_left;
- stereo->smooth_right = .98*stereo->smooth_right + .02*e_right;
- data[2*i] = stereo->smooth_left*ftmp;
- data[2*i+1] = stereo->smooth_right*ftmp;
+ spx_int16_t tmp=data[i];
+ stereo->smooth_left = EXTRACT16(PSHR32(MAC16_16(MULT16_16(stereo->smooth_left, QCONST16(0.98, 15)), e_left, QCONST16(0.02, 15)), 15));
+ stereo->smooth_right = EXTRACT16(PSHR32(MAC16_16(MULT16_16(stereo->smooth_right, QCONST16(0.98, 15)), e_right, QCONST16(0.02, 15)), 15));
+ data[2*i] = (spx_int16_t)MULT16_16_P14(stereo->smooth_left, tmp);
+ data[2*i+1] = (spx_int16_t)MULT16_16_P14(stereo->smooth_right, tmp);
}
}
int speex_std_stereo_request_handler(SpeexBits *bits, void *state, void *data)
{
- SpeexStereoState *stereo;
- float sign=1;
+ RealSpeexStereoState *stereo;
+ spx_word16_t sign=1, dexp;
int tmp;
- stereo = (SpeexStereoState*)data;
+ stereo = (RealSpeexStereoState*)data;
+
+ COMPATIBILITY_HACK(stereo);
+
if (speex_bits_unpack_unsigned(bits, 1))
sign=-1;
- tmp = speex_bits_unpack_unsigned(bits, 5);
- stereo->balance = exp(sign*.25*tmp);
-
+ dexp = speex_bits_unpack_unsigned(bits, 5);
+#ifndef FIXED_POINT
+ stereo->balance = exp(sign*.25*dexp);
+#else
+ stereo->balance = spx_exp(MULT16_16(sign, SHL16(dexp, 9)));
+#endif
tmp = speex_bits_unpack_unsigned(bits, 2);
stereo->e_ratio = e_ratio_quant[tmp];
Modified: trunk/speex/libspeex/vq.c
===================================================================
--- trunk/speex/libspeex/vq.c 2007-11-13 15:42:33 UTC (rev 14136)
+++ trunk/speex/libspeex/vq.c 2007-11-14 11:59:22 UTC (rev 14137)
@@ -70,30 +70,7 @@
return i;
}
-/*Finds the index of the entry in a codebook that best matches the input*/
-int vq_index(float *in, const float *codebook, int len, int entries)
-{
- int i,j;
- float min_dist=0;
- int best_index=0;
- for (i=0;i<entries;i++)
- {
- float dist=0;
- for (j=0;j<len;j++)
- {
- float tmp = in[j]-*codebook++;
- dist += tmp*tmp;
- }
- if (i==0 || dist<min_dist)
- {
- min_dist=dist;
- best_index=i;
- }
- }
- return best_index;
-}
-
#ifndef OVERRIDE_VQ_NBEST
/*Finds the indices of the n-best entries in a codebook*/
void vq_nbest(spx_word16_t *in, const spx_word16_t *codebook, int len, int entries, spx_word32_t *E, int N, int *nbest, spx_word32_t *best_dist, char *stack)
Modified: trunk/speex/libspeex/vq.h
===================================================================
--- trunk/speex/libspeex/vq.h 2007-11-13 15:42:33 UTC (rev 14136)
+++ trunk/speex/libspeex/vq.h 2007-11-14 11:59:22 UTC (rev 14137)
@@ -40,7 +40,6 @@
int scal_quant(spx_word16_t in, const spx_word16_t *boundary, int entries);
int scal_quant32(spx_word32_t in, const spx_word32_t *boundary, int entries);
-int vq_index(float *in, const float *codebook, int len, int entries);
#ifdef _USE_SSE
#include <xmmintrin.h>
void vq_nbest(spx_word16_t *in, const __m128 *codebook, int len, int entries, __m128 *E, int N, int *nbest, spx_word32_t *best_dist, char *stack);
More information about the commits
mailing list