shithub: opus

Download patch

ref: c5f2a9d988fe4fddae0b499d1a12d12cee6c849f
parent: d77f61ac22fa57d2f24039a959ea722b9d725629
author: Jean-Marc Valin <[email protected]>
date: Sun Oct 26 18:00:26 EDT 2008 

Latest psychoacoustics work -- still highly experimental


--- a/libcelt/celt.c
+++ b/libcelt/celt.c
@@ -390,6 +390,7 @@
    VARDECL(int, offsets);
 #ifdef EXP_PSY
    VARDECL(celt_word32_t, mask);
+   VARDECL(celt_word32_t, tonality);
 #endif
    int shortBlocks=0;
    int transient_time;
@@ -468,18 +469,22 @@
       transient_shift = 0;
       shortBlocks = 0;
    }
+
    /* Pitch analysis: we do it early to save on the peak stack space */
-   if (st->pitch_enabled && !shortBlocks)
-   {
 #ifdef EXP_PSY
+   ALLOC(tonality, MAX_PERIOD/4, celt_word16_t);
+   {
       VARDECL(celt_word16_t, X);
       ALLOC(X, MAX_PERIOD/2, celt_word16_t);
       find_spectral_pitch(st->mode, st->mode->fft, &st->mode->psy, in, st->out_mem, st->mode->window, X, 2*N-2*N4, MAX_PERIOD-(2*N-2*N4), &pitch_index);
-      compute_tonality(st->mode, X, st->psy_mem, MAX_PERIOD);
+      compute_tonality(st->mode, X, st->psy_mem, MAX_PERIOD, tonality, MAX_PERIOD/4);
+   }
 #else
+   if (st->pitch_enabled && !shortBlocks)
+   {
       find_spectral_pitch(st->mode, st->mode->fft, &st->mode->psy, in, st->out_mem, st->mode->window, NULL, 2*N-2*N4, MAX_PERIOD-(2*N-2*N4), &pitch_index);
-#endif
    }
+#endif
    ALLOC(freq, C*N, celt_sig_t); /**< Interleaved signal MDCTs */
    
    /*for (i=0;i<(B+1)*C*N;i++) printf ("%f(%d) ", in[i], i); printf ("\n");*/
@@ -495,13 +500,16 @@
          st->psy_mem[MAX_PERIOD+st->overlap-N+i] += in[C*(st->overlap+i)+c];
    */
    ALLOC(mask, N, celt_sig_t);
-   compute_mdct_masking(&st->psy, freq, st->psy_mem, mask, C*N);
+   compute_mdct_masking(&st->psy, freq, tonality, st->psy_mem, mask, C*N);
+   /*for (i=0;i<256;i++)
+      printf ("%f %f %f ", freq[i], tonality[i], mask[i]);
+   printf ("\n");*/
 
    /* Invert and stretch the mask to length of X 
       For some reason, I get better results by using the sqrt instead,
       although there's no valid reason to. Must investigate further */
-   for (i=0;i<C*N;i++)
-      mask[i] = 1/(.1+mask[i]);
+   /*for (i=0;i<C*N;i++)
+      mask[i] = 1/(.1+mask[i]);*/
 #endif
    
    /* Deferred allocation after find_spectral_pitch() to reduce the peak memory usage */
@@ -520,6 +528,16 @@
 
    /* Band normalisation */
    compute_band_energies(st->mode, freq, bandE);
+#ifdef EXP_PSY
+   VARDECL(celt_word32_t, bandM);
+   ALLOC(bandM,st->mode->nbEBands, celt_ener_t);
+   for (i=0;i<N;i++)
+      mask[i] = sqrt(mask[i]);
+   compute_band_energies(st->mode, mask, bandM);
+   /*for (i=0;i<st->mode->nbEBands;i++)
+      printf ("%f %f ", bandE[i], bandM[i]);
+   printf ("\n");*/
+#endif
    normalise_bands(st->mode, freq, X, bandE);
    /*for (i=0;i<st->mode->nbEBands;i++)printf("%f ", bandE[i]);printf("\n");*/
    /*for (i=0;i<N*B*C;i++)printf("%f ", X[i]);printf("\n");*/
--- a/libcelt/psy.c
+++ b/libcelt/psy.c
@@ -145,7 +145,7 @@
 }
 
 #ifdef EXP_PSY /* Not needed for now, but will be useful in the future */
-void compute_mdct_masking(const struct PsyDecay *decay, celt_word32_t *X, celt_word16_t *long_window, celt_mask_t *mask, int len)
+void compute_mdct_masking(const struct PsyDecay *decay, celt_word32_t *X, celt_word16_t *tonality, celt_word16_t *long_window, celt_mask_t *mask, int len)
 {
    int i;
    VARDECL(float, psd);
@@ -152,7 +152,7 @@
    SAVE_STACK;
    ALLOC(psd, len, float);
    for (i=0;i<len;i++)
-      psd[i] = X[i]*X[i];
+      psd[i] = X[i]*X[i]*tonality[i];
    for (i=1;i<len-1;i++)
       mask[i] = .5*psd[i] + .25*(psd[i-1]+psd[i+1]);
    /*psd[0] = .5*mask[0]+.25*(mask[1]+mask[2]);*/
@@ -164,7 +164,7 @@
    RESTORE_STACK;  
 }
 
-void compute_tonality(const CELTMode *m, celt_word16_t * restrict X, celt_word16_t * mem, int len)
+void compute_tonality(const CELTMode *m, celt_word16_t * restrict X, celt_word16_t * mem, int len, celt_word16_t *tonality, int mdct_size)
 {
    int i;
    celt_word16_t norm_1;
@@ -174,6 +174,7 @@
    mem2 = mem+2*N;
    X[0] = 0;
    X[1] = 0;
+   tonality[0] = 1;
    for (i=1;i<N;i++)
    {
       celt_word16_t re, im, re2, im2;
@@ -180,7 +181,7 @@
       re = X[2*i];
       im = X[2*i+1];
       /* Normalise spectrum */
-      norm_1 = celt_rsqrt(MAC16_16(MULT16_16(re,re), im,im));
+      norm_1 = celt_rsqrt(.01+MAC16_16(MULT16_16(re,re), im,im));
       re = MULT16_16(re, norm_1);
       im = MULT16_16(im, norm_1);
       /* Phase derivative */
@@ -198,5 +199,13 @@
       X[2*i+1] = im;
    }
    /*printf ("\n");*/
+   for (i=0;i<mdct_size;i++)
+   {
+      tonality[i] = 1.0-X[2*i]*X[2*i]*X[2*i];
+      if (tonality[i]>1)
+         tonality[i] = 1;
+      if (tonality[i]<.02)
+         tonality[i]=.02;
+   }
 }
 #endif
--- a/libcelt/psy.h
+++ b/libcelt/psy.h
@@ -49,8 +49,8 @@
 void compute_masking(const struct PsyDecay *decay, celt_word16_t *X, celt_mask_t *mask, int len);
 
 /** Compute the masking curve for an input (MDCT) spectrum X */
-void compute_mdct_masking(const struct PsyDecay *decay, celt_word32_t *X, celt_word16_t *long_window, celt_mask_t *mask, int len);
+void compute_mdct_masking(const struct PsyDecay *decay, celt_word32_t *X, celt_word16_t *tonality, celt_word16_t *long_window, celt_mask_t *mask, int len);
 
-void compute_tonality(const CELTMode *m, celt_word16_t * restrict X, celt_word16_t * mem, int len);
+void compute_tonality(const CELTMode *m, celt_word16_t * restrict X, celt_word16_t * mem, int len, celt_word16_t *tonality, int mdct_size);
 
 #endif /* PSY_H */