new filter subdir for standard audio filters first filter uses code from vorbis to implement an iir filter not optimi...

Original commit message from CVS:
new filter subdir for standard audio filters
first filter uses code from vorbis to implement an iir filter
not optimized yet, iir code uses doubles and plugin uses float

1 files changed, 302 insertions, 0 deletions

diff --git a/gst/filter/iir.c b/gst/filter/iir.c
new file mode 100644
index 00000000..1ef31540
--- /dev/null
+++ b/gst/filter/iir.c
@@ -0,0 +1,302 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE Ogg Vorbis SOFTWARE CODEC SOURCE CODE.  *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS SOURCE IS GOVERNED BY *
+ * THE GNU PUBLIC LICENSE 2, WHICH IS INCLUDED WITH THIS SOURCE.    *
+ * PLEASE READ THESE TERMS DISTRIBUTING.                            *
+ *                                                                  *
+ * THE OggSQUISH SOURCE CODE IS (C) COPYRIGHT 1994-2000             *
+ * by Monty <monty@xiph.org> and The XIPHOPHORUS Company            *
+ * http://www.xiph.org/                                             *
+ *                                                                  *
+ ********************************************************************
+
+  function: Direct Form I, II IIR filters, plus some specializations
+  last mod: $Id$
+
+ ********************************************************************/
+
+/* LPC is actually a degenerate case of form I/II filters, but we need
+   both */
+
+#include <stdlib.h>
+#include <string.h>
+#include <math.h>
+#include "iir.h"
+
+void IIR_init(IIR_state *s,int stages,double gain, double *A, double *B){
+  memset(s,0,sizeof(IIR_state));
+  s->stages=stages;
+  s->gain=gain;
+  s->coeff_A=malloc(stages*sizeof(double));
+  s->coeff_B=malloc((stages+1)*sizeof(double));
+  s->z_A=calloc(stages*2,sizeof(double));
+  s->z_B=calloc(stages*2,sizeof(double));
+
+  memcpy(s->coeff_A,A,stages*sizeof(double));
+  memcpy(s->coeff_B,B,(stages+1)*sizeof(double));
+}
+
+void IIR_clear(IIR_state *s){
+  if(s){
+    free(s->coeff_A);
+    free(s->coeff_B);
+    free(s->z_A);
+    free(s->z_B);
+    memset(s,0,sizeof(IIR_state));
+  }
+}
+
+double IIR_filter(IIR_state *s,double in){
+  int stages=s->stages,i;
+  double newA;
+  double newB=0;
+  double *zA=s->z_A+s->ring;
+
+  newA=in/=s->gain;
+  for(i=0;i<stages;i++){
+    newA+= s->coeff_A[i] * zA[i];
+    newB+= s->coeff_B[i] * zA[i];
+  }
+  newB+=newA*s->coeff_B[stages];
+
+  zA[0]=zA[stages]=newA;
+  if(++s->ring>=stages)s->ring=0;
+
+  return(newB);
+}
+
+/* this assumes the symmetrical structure of the feed-forward stage of
+   a Chebyshev bandpass to save multiplies */
+double IIR_filter_ChebBand(IIR_state *s,double in){
+  int stages=s->stages,i;
+  double newA;
+  double newB=0;
+  double *zA=s->z_A+s->ring;
+
+  newA=in/=s->gain;
+
+  newA+= s->coeff_A[0] * zA[0];
+  for(i=1;i<(stages>>1);i++){
+    newA+= s->coeff_A[i] * zA[i];
+    newB+= s->coeff_B[i] * (zA[i]-zA[stages-i]);
+  }
+  newB+= s->coeff_B[i] * zA[i];
+  for(;i<stages;i++)
+    newA+= s->coeff_A[i] * zA[i];
+
+  newB+= newA-zA[0];
+
+  zA[0]=zA[stages]=newA;
+  if(++s->ring>=stages)s->ring=0;
+
+  return(newB);
+}
+
+#ifdef _V_SELFTEST
+
+/* z^-stage, z^-stage+1... */
+static double cheb_bandpass_B[]={-1.,0.,5.,0.,-10.,0.,10.,0.,-5.,0.,1};
+static double cheb_bandpass_A[]={-0.6665900311,
+				  1.0070146601,
+				 -3.1262875409,
+			 	  3.5017171569,
+				 -6.2779211945,
+				  5.2966481740,
+				 -6.7570216587,
+				  4.0760335768,
+				 -3.9134284363,
+				  1.3997338886};
+
+static double data[128]={  
+  0.0426331,
+  0.0384521,
+  0.0345764,
+  0.0346069,
+  0.0314636,
+  0.0310059,
+  0.0318604,
+  0.0336304,
+  0.036438,
+  0.0348511,
+  0.0354919,
+  0.0343628,
+  0.0325623,
+  0.0318909,
+  0.0263367,
+  0.0225525,
+  0.0195618,
+  0.0160828,
+  0.0168762,
+  0.0145569,
+  0.0126343,
+  0.0127258,
+  0.00820923,
+  0.00787354,
+  0.00558472,
+  0.00204468,
+  3.05176e-05,
+  -0.00357056,
+  -0.00570679,
+  -0.00991821,
+  -0.0101013,
+  -0.00881958,
+  -0.0108948,
+  -0.0110168,
+  -0.0119324,
+  -0.0161438,
+  -0.0194702,
+  -0.0229187,
+  -0.0260315,
+  -0.0282288,
+  -0.0306091,
+  -0.0330505,
+  -0.0364685,
+  -0.0385742,
+  -0.0428772,
+  -0.043457,
+  -0.0425415,
+  -0.0462341,
+  -0.0467529,
+  -0.0489807,
+  -0.0520325,
+  -0.0558167,
+  -0.0596924,
+  -0.0591431,
+  -0.0612793,
+  -0.0618591,
+  -0.0615845,
+  -0.0634155,
+  -0.0639648,
+  -0.0683594,
+  -0.0718079,
+  -0.0729675,
+  -0.0791931,
+  -0.0860901,
+  -0.0885315,
+  -0.088623,
+  -0.089386,
+  -0.0899353,
+  -0.0886841,
+  -0.0910645,
+  -0.0948181,
+  -0.0919495,
+  -0.0891418,
+  -0.0916443,
+  -0.096344,
+  -0.100464,
+  -0.105499,
+  -0.108612,
+  -0.112213,
+  -0.117676,
+  -0.120911,
+  -0.124329,
+  -0.122162,
+  -0.120605,
+  -0.12326,
+  -0.12619,
+  -0.128998,
+  -0.13205,
+  -0.134247,
+  -0.137939,
+  -0.143555,
+  -0.14389,
+  -0.14859,
+  -0.153717,
+  -0.159851,
+  -0.164551,
+  -0.162811,
+  -0.164276,
+  -0.156952,
+  -0.140564,
+  -0.123291,
+  -0.10321,
+  -0.0827637,
+  -0.0652466,
+  -0.053772,
+  -0.0509949,
+  -0.0577698,
+  -0.0818176,
+  -0.114929,
+  -0.148895,
+  -0.181122,
+  -0.200714,
+  -0.21048,
+  -0.203644,
+  -0.179413,
+  -0.145325,
+  -0.104492,
+  -0.0658264,
+  -0.0332031,
+  -0.0106201,
+  -0.00363159,
+  -0.00909424,
+  -0.0244141,
+  -0.0422058,
+  -0.0537415,
+  -0.0610046,
+  -0.0609741,
+  -0.0547791};
+
+/* comparison test code from http://www-users.cs.york.ac.uk/~fisher/mkfilter/
+   (the above page kicks ass, BTW)*/
+
+#define NZEROS 10
+#define NPOLES 10
+#define GAIN   4.599477515e+02
+
+static float xv[NZEROS+1], yv[NPOLES+1];
+
+static double filterloop(double next){ 
+  xv[0] = xv[1]; xv[1] = xv[2]; xv[2] = xv[3]; xv[3] = xv[4]; xv[4] = xv[5]; 
+  xv[5] = xv[6]; xv[6] = xv[7]; xv[7] = xv[8]; xv[8] = xv[9]; xv[9] = xv[10]; 
+  xv[10] = next / GAIN;
+  yv[0] = yv[1]; yv[1] = yv[2]; yv[2] = yv[3]; yv[3] = yv[4]; yv[4] = yv[5]; 
+  yv[5] = yv[6]; yv[6] = yv[7]; yv[7] = yv[8]; yv[8] = yv[9]; yv[9] = yv[10]; 
+  yv[10] =   (xv[10] - xv[0]) + 5 * (xv[2] - xv[8]) + 10 * (xv[6] - xv[4])
+    + ( -0.6665900311 * yv[0]) + (  1.0070146601 * yv[1])
+    + ( -3.1262875409 * yv[2]) + (  3.5017171569 * yv[3])
+    + ( -6.2779211945 * yv[4]) + (  5.2966481740 * yv[5])
+    + ( -6.7570216587 * yv[6]) + (  4.0760335768 * yv[7])
+    + ( -3.9134284363 * yv[8]) + (  1.3997338886 * yv[9]);
+  return(yv[10]);
+}
+
+#include <stdio.h>
+int main(){
+
+  /* run the pregenerated Chebyshev filter, then our own distillation
+     through the generic and specialized code */
+  double *work=malloc(128*sizeof(double));
+  IIR_state iir;
+  int i;
+
+  for(i=0;i<128;i++)work[i]=filterloop(data[i]);
+  {
+    FILE *out=fopen("IIR_ref.m","w");
+    for(i=0;i<128;i++)fprintf(out,"%g\n",work[i]);
+    fclose(out);
+  }
+
+  IIR_init(&iir,NPOLES,GAIN,cheb_bandpass_A,cheb_bandpass_B);
+  for(i=0;i<128;i++)work[i]=IIR_filter(&iir,data[i]);
+  {
+    FILE *out=fopen("IIR_gen.m","w");
+    for(i=0;i<128;i++)fprintf(out,"%g\n",work[i]);
+    fclose(out);
+  }
+  IIR_clear(&iir);  
+
+  IIR_init(&iir,NPOLES,GAIN,cheb_bandpass_A,cheb_bandpass_B);
+  for(i=0;i<128;i++)work[i]=IIR_filter_ChebBand(&iir,data[i]);
+  {
+    FILE *out=fopen("IIR_cheb.m","w");
+    for(i=0;i<128;i++)fprintf(out,"%g\n",work[i]);
+    fclose(out);
+  }
+  IIR_clear(&iir);  
+
+  return(0);
+}
+
+#endif