Pulse analysis using ordinal value filtering

Boots – shoes – and leggings

Patent

Rate now

  [ 0.00 ] – not rated yet Voters 0   Comments 0

Details

382218, 382262, G06F 1710

Patent

active

059177330

DESCRIPTION:

BRIEF SUMMARY
FIELD OF THE INVENTION

This invention relates generally to signal processing and more specifically to methods and apparatus for noise reduction, signal analysis, pattern recognition and data compression.


BACKGROUND TO THE INVENTION

The invention makes use of ordinal value filters, which term is to be interpreted in the light of subsequent comments, and a brief explanation of the art relating to these filters is appropriate before describing the invention. An alternative term for ordinal value filters, sometimes used in relevant literature, is a rank filter.
A real time ordinal value filter which determines which one of a set of R applied data values is the Nth-largest, becomes a median filter if R is odd and N is made equal to (R+1)/2. In such a mode the filter provides as an output the "middle" value of the data set, having in general an equal number of other data values from the set larger and smaller than itself.
The design and operation of median filters is described inter alia in U.S. Pat. Nos. 4,441,165, 4,439,840, 4,560,974 and 45,135,440.
In a simple case of a linear array of data samples, if R=3 and N=2, the median value is the middle value of each three successive data values when the latter are sorted into order, beginning with the smallest and ending with the largest of each group of three. Since there is no averaging the median filter will remove spurious samples without degrading sharp transitions in the original sample series. Put another way, if the data samples represent a regularly sampled, time dependent signal, then the median filter has the property of not restricting the band width of the signal (which would degrade steep transitions in the original signal) but instead allows original signal transitions, representing long duration features, to be transmitted at full band width while substantially attenuating shorter duration spurious spikes, characteristic of electrical noise.
The particular threshold at which such a filter will eliminate spurious signals but pass other signals is determined by the values of R and N. The greater the number of atypical samples that represent the spurious signal that is to be eliminated, the greater will have to be the value of R.
However, it has been found that a median filter, set to eliminate spurious spikes represented by less than N samples for example, may actually introduce spurious signals that are longer than N samples in duration into the filter output, when a succession of short spikes are closer together than N sampling intervals. This can have serious repercussions if the information is contained, at least in part, in the number of samples representing the amplitude excursions in the original.
In an attempt to further enhance noise reduction, it has been proposed to repeat a median filtering step by applying the filtered signal from a first median filter, either to the input of the same filter again or to a second identical filter. If this is repeated until no further changes are introduced by the signal, it is called a root median filter. At first it was thought that this might overcome the introduction of spurious signals as discussed above, but it has been found that this repeated ordinal value filtering does not resolve the problem.
Experiments have shown that the introduction of adventitious information into the filter output is reduced when the value of R is small. For example, in the case of a linear sequence of samples, if R=3, it would appear that there is little likelihood of any spurious additional signals appearing in the output. By appropriate choice of sampling interval, a median filter with a value of R=3 should not introduce any unwanted signal transitions in the filter output signal. However, spurious signals represented by two or more samples will not be suppressed, which, when the sampling rate is high, may not result in very satisfactory noise reduction.
Increasing the number of samples representing a given signal will not necesssarily result in an increase in the accuracy of the filtered signal since the value of R

REFERENCES:
patent: 4603430 (1986-07-01), Sacks
patent: 4713786 (1987-12-01), Roskind
patent: 4970637 (1990-11-01), Sato
patent: 5097433 (1992-03-01), Caracciolo
patent: 5319583 (1994-06-01), Wildes

LandOfFree

Say what you really think

Search LandOfFree.com for the USA inventors and patents. Rate them and share your experience with other people.

Rating

Pulse analysis using ordinal value filtering does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Pulse analysis using ordinal value filtering, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Pulse analysis using ordinal value filtering will most certainly appreciate the feedback.

Rate now

     

Profile ID: LFUS-PAI-O-1381651

  Search
All data on this website is collected from public sources. Our data reflects the most accurate information available at the time of publication.