summaryrefslogtreecommitdiff
path: root/src/c/signalProcessing/wiener/dwienera.c
diff options
context:
space:
mode:
Diffstat (limited to 'src/c/signalProcessing/wiener/dwienera.c')
-rw-r--r--src/c/signalProcessing/wiener/dwienera.c166
1 files changed, 166 insertions, 0 deletions
diff --git a/src/c/signalProcessing/wiener/dwienera.c b/src/c/signalProcessing/wiener/dwienera.c
new file mode 100644
index 0000000..18f49bc
--- /dev/null
+++ b/src/c/signalProcessing/wiener/dwienera.c
@@ -0,0 +1,166 @@
+/* Copyright (C) 2017 - IIT Bombay - FOSSEE
+
+ This file must be used under the terms of the CeCILL.
+ This source file is licensed as described in the file COPYING, which
+ you should have received as part of this distribution. The terms
+ are also available at
+ http://www.cecill.info/licences/Licence_CeCILL_V2-en.txt
+ Author: Abhinav Dronamraju
+ Organization: FOSSEE, IIT Bombay
+ Email: toolbox@scilab.in
+ */
+
+/*Function to find kalm */
+
+#include "lapack.h"
+#include "stdlib.h"
+#include "stdio.h"
+#include "string.h"
+#include "wiener.h"
+#include "kalm.h"
+#include "matrixTranspose.h"
+#include "matrixMultiplication.h"
+#include "matrixInversion.h"
+#include "addition.h"
+#include "subtraction.h"
+#include "eye.h"
+
+void dwienera(double* y, int y_row, int y_col, double* x0, int x0_row, int x0_col, double* p0, int p0_row, int p0_col, double* f, int f_row, int f_col, double* g, int g_row, int g_col, double* h, int h_row, int h_col, double* q, int q_row, int q_col, double* r, int r_row, int r_col, double* xs, double* ps, double* xf, double* pf)
+
+{
+
+/* dtransposea(h, h_row, h_col, h_trans);*/
+/* dmulma(p0, p0_row, p0_col, h_trans, h_col, h_row, temp1); //temp1= p0*h'*/
+/* dmulma(h, h_row, h_col, p0, p0_row, p0_col, temp2); //temp2= h*p0*/
+/* dmulma(temp2, h_row, p0_col, h_trans, h_col, h_row, temp3); //temp3= h*p0*h'*/
+/* dadda(temp3, h_row*h_row, r, r_row*r_col, temp4); //temp4= h*p0*h'+r*/
+
+ int n= x0_row, x0j= x0_col, m= y_row, tf= y_col, to=1, k;
+ double ind_nk[n];
+ double ind_mk[m];
+ double xf1[x0_row*x0_col];
+ double pf1[p0_row*p0_col];
+ double yk[y_row*1];
+ double fk[f_row*n];
+ double gk[g_row*n];
+ double hk[h_row*n];
+ double qk[q_row*n];
+ double rk[r_row*m];
+ double x1[f_row*x0_col];
+ double p1[f_row*f_row];
+ double x[x0_row*x0_col];
+ double p[p0_row*p0_col];
+
+
+for(int i=0; i< x0_row*x0_col; i++)
+ xf1[i]= x0[i];
+
+for(int i=0; i< p0_row*p0_col; i++)
+ pf1[i]= p0[i];
+
+for(k=t0; k<=tf; k++)
+{
+ int j=0;
+ for(int i=1+(k-1)*n; i<= k*n; i++)
+ {
+ ind_nk[j]= i;
+ j++;
+ }
+
+ int j=0;
+ for(int i=1+(k-1)*m; i<= k*m; i++)
+ {
+ ind_mk[j]= i;
+ j++;
+ }
+
+ for(int i=0; i< y_row; y++)
+ {
+ int j=k-1;
+
+ yk[i]= y[i+j*y_row];
+
+ }
+int l=0;
+ for(int i=0; i<n; i++)
+ {
+ for(int j=0; j< f_row; j++)
+ {
+ int k=ind_nk[i]-1;
+
+ fk[l]= f[j+k*y_row];
+ l++;
+ }
+
+
+ }
+
+int l=0;
+ for(int i=0; i<n; i++)
+ {
+ for(int j=0; j< g_row; j++)
+ {
+ int k=ind_nk[i]-1;
+
+ gk[l]= g[j+k*y_row];
+ l++;
+ }
+
+
+ }
+
+int l=0;
+ for(int i=0; i<n; i++)
+ {
+ for(int j=0; j< h_row; j++)
+ {
+ int k=ind_nk[i]-1;
+
+ hk[l]= h[j+k*y_row];
+ l++;
+ }
+
+
+ }
+
+int l=0;
+ for(int i=0; i<n; i++)
+ {
+ for(int j=0; j< q_row; j++)
+ {
+ int k=ind_nk[i]-1;
+
+ qk[l]= q[j+k*y_row];
+ l++;
+ }
+
+
+ }
+
+int l=0;
+ for(int i=0; i<m; i++)
+ {
+ for(int j=0; j< r_row; j++)
+ {
+ int k=ind_mk[i]-1;
+
+ rk[l]= r[j+k*y_row];
+ l++;
+ }
+
+
+ }
+
+
+dkalma(yk, y_row, 1, x0, x0_row, x0_col, p0, p0_row, p0_col, fk, f_row, n, gk, g_row, n, hk, h_row, n, qk, q_row, n, rk, r_row, m, x1, p1, x, p);
+
+
+
+
+}
+
+
+
+
+
+}