can anybody let me know what is pseudo random sub sampling used in pattern matching

[Shreyas]

Hi,

     If you sub-sample the template by taking a uniform grid, it is possible to miss important features of the template, such as horizontal and vertical edges. Random sampling tends to produce clusters and open areas; having clusters of samples in the same area of the template is inefficient, while the unsampled areas may contain important information. The pseudo-random technique maximizes the uniformity of the sampling throughout the template without using a pre-defined grid.

For more information go to Start>> programs>>National Instruments>>Vision>>Documentation. The PDF's contain lot of information on pattern matching and sampling.

Also refer : http://zone.ni.com/devzone/cda/tut/p/id/3763

Hope that helps you build your application.

Kudos are welcome.

[muks]

have you seen this?

[muks]

ok shreyas had given the link. (Too much of time looking at the content cost me.........)

[raviraju]

dear sir,

i have already seen the link. but it is not clear how uniform sampling misses the important features and pseudo random subsampling does not miss that information.

the topic needs more elaboration. pl. help.

thans 

[Shreyas]

Hi,

     When multiple random test-and-train experiments are performed, a new classifier is learned from each training sample. The estimated error rate is the average of the error rates for classifiers derived for the independently and randomly generated test partitions. Random subsampling can produce better error estimates than a single train-and-test partition.

In sub-sampling and sample rate decimating discrete time systems it is necessary to suppress interferers whose frequencies fall within the aliasing bandwidths of the particular system. For this reason, anti-aliasing filters are used to suppressthese interferers. The requirements of these anti-aliasing filters, however, are typically very demanding due to the level of interference encountered, bandwidths of the desired and interfering signals, desired suppression, etc. The demandingrequirements of the anti-aliasing filter typically results in having to design in costly analog filters or high gate count digital filters. In some cases, it is too cost prohibitive to include an anti-aliasing filter in the system and performance istherefore compromised, for example, in very low cost consumer applications.

A requirement, however, is that the sampling frequency always remain above at least twice the maximum frequency of the desired signal. If this is maintained, the desired signal will not be affected by the constantly changing sampling frequency. The interfering signal, however, will be spread across the spectrum due to the aliasing affect of the lower sampling frequency. The range of sampling frequencies should be selected such that the average sampling frequency is equal to the desiredsampling rate. Consider sampling at a frequency of 2f₁ during one cycle and then sampling at a frequency of 2f_1+Δ at the next cycle. The sequence and actual sampling frequencies are not critical as long as the average samplingfrequency is maintained at the desired sampling rate. The sampling frequency is randomized over time symmetrically around a mean value. It is ensured that the lower end of the range of sampling frequencies is higher than twice the maximum frequency ofthe desired signal.

Kudos are welcome