Performance of TSA toolbox
 
Very efficient algorithms have been used, as in the following three examples can be seen:

AR-test, calculates Autoregressive parameters with the Yule-Walker method
d:\programs\matlab\toolbox\ident\ar.m
   needs  628677 FLOPS,0.140s for 1024 samples and a model order of 10.
s:\matlab\dpmi\tsa\earpyw.m
   needs  22812 FLOPS, 0.050s for 1024 samples and a model order of 10.

Histogram test
d:\programs\matlab\toolbox\matlab\datafun\hist.m
   needs  1089687 FLOPS, 1.592s for 1024 samples and a binwidth of 0.01 (1060 bins).
s:\matlab\dpmi\tsa\histo.m
   needs  5143 FLOPS, 0.150s for 1024 samples and a binwidth of 0.01 (1060 bins).

UCP-Test: Test polynomials if all roots are inside the Unit Circle
for k=1:N, b(k)=all(abs(roots(y(k,:)))<1); end;
   needs  1864656 FLOPS, 0.591s for 100 polynomials of order 11.
s:\matlab\dpmi\tsa\ucp.m
   needs  12040 FLOPS, 0.030s for 100 polynomials of order 11.

The results were obtained on a Intel Pentium 166MHz and Matlab 5.1. You can can get the results for Your platform by running demoperf.m of the TSA toolbox (V2.44 or higher) on your platform.


Back