Method and apparatus for performing pattern recognition analysis
First Claim
1. A method of performing pattern analysis of a digital signal containing at least one waveform, comprising the steps of:
- detecting a fiducial point of each waveform of said digital signal, said step of detecting said fiducial point of each waveform including the step of detecting a segment of the respective waveform, which segment has a slope with an absolute value greater than a trigger value, said step of detecting a segment including the steps of forming a delta value for each sampling point of said digital signal by subtracting the value of each sampling point from the value of the sampling point occurring m sampling points previous thereto, determining the average of n highest differences in values between consecutive sampling points in a predetermined interval of said digital signal, obtaining said trigger value based on said average, and detecting a first sampling point having a delta value at least equal to said trigger value as said fiducial point, where m an n are integers;
capturing at least two characteristic features of each waveform of said digital signal;
creating a signature for each waveform based on said captured at least two characteristic features of each waveform;
determining a closest proximity of said signature to one classification of at least one previously determined classification of signatures;
creating a new classification with said signature if said closest proximity has a value less than a predetermined threshold value; and
adding said signature to said one classification if said closest proximity has a value at least equal to said predetermined threshold value.
7 Assignments
0 Petitions
Accused Products
Abstract
A method and apparatus for performing pattern analysis of a digital signal containing at least one waveform, detects a fiducial point of each waveform of the digital signal; captures slope transition features separated by interval features of each of the waveforms of the digital signal; creates a signature for each waveform based on the captured slope transition and interval features of each of said waveforms; determines the closest proximity of the signature to one of a plurality of previously determined classifications of signatures; creates a new classification with the signature if the closest proximity has a value less than a predetermined threshold value; and adds the signature to the closest classification if the closest proximity has a value at least equal to the predetermined threshold value.
54 Citations
24 Claims
-
1. A method of performing pattern analysis of a digital signal containing at least one waveform, comprising the steps of:
-
detecting a fiducial point of each waveform of said digital signal, said step of detecting said fiducial point of each waveform including the step of detecting a segment of the respective waveform, which segment has a slope with an absolute value greater than a trigger value, said step of detecting a segment including the steps of forming a delta value for each sampling point of said digital signal by subtracting the value of each sampling point from the value of the sampling point occurring m sampling points previous thereto, determining the average of n highest differences in values between consecutive sampling points in a predetermined interval of said digital signal, obtaining said trigger value based on said average, and detecting a first sampling point having a delta value at least equal to said trigger value as said fiducial point, where m an n are integers; capturing at least two characteristic features of each waveform of said digital signal; creating a signature for each waveform based on said captured at least two characteristic features of each waveform; determining a closest proximity of said signature to one classification of at least one previously determined classification of signatures; creating a new classification with said signature if said closest proximity has a value less than a predetermined threshold value; and adding said signature to said one classification if said closest proximity has a value at least equal to said predetermined threshold value. - View Dependent Claims (2, 3)
-
-
4. A method of performing pattern analysis of a digital signal containing at least one waveform, comprising the steps of:
-
detecting a fiducial point of each waveform of said digital signal, said step of detecting said fiducial point of each waveform including the step of detecting a segment of the respective waveform, which segment has a slope with an absolute value greater than a trigger value, capturing at least two characteristic features of each waveform of said digital signal, said step of capturing includes the step of capturing a plurality of interval features and a plurality of slope transition features separated by said plurality of interval features, said step of capturing a plurality of interval features and a plurality of slope transition features separated by said plurality of interval features includes the step of detecting each of a plurality of slope transitions between a positive slope, a negative slope and a flat, each slope transition separated by a plurality of intervals, and said step of detecting each of a plurality of slope transitions separated by a plurality of intervals includes the steps of forming a delta value for each sampling point by subtracting a value of each sampling point from a value of the sampling point occurring m sampling points previous thereto, determining the average of n highest differences in values between consecutive sampling points in a predetermined interval of said digital signal, obtaining a trigger value based on said average, and determining said plurality of slope transitions and the plurality of intervals in accordance with a relation between the delta values for said sampling points and said trigger value where m and n are integers; creating a signature for each waveform based on said captured at least two characteristic features of each waveform; determining a closest proximity of said signature to one classification of at least one previously determined classification of signatures; creating a new classification with said signature if said closest proximity has a value less than a predetermined threshold value; and adding said signature to said one classification if said closest proximity has a value at least equal to said predetermined threshold value. - View Dependent Claims (5, 6, 7)
-
-
8. A method of performing pattern analysis of a digital signal containing at least one waveform, comprising the steps of:
-
detecting a fiducial point of each waveform of said digital signal, said step of detecting said fiducial point of each waveform including the step of detecting a segment of respective waveform, which segment has a slope with an absolute value greater than a trigger value, capturing at least two characteristic features of each waveform of said digital signal, said step of capturing includes the step of capturing a plurality of slope transition features and a plurality of interval features that separate said slope transition features; creating a signature for each waveform based on said captured at least two characteristic features of each waveform;
said step of creating said signature includes the step of assigning different values to each of said captured slope transition features and interval features;determining a closest proximity of said signature to one classification of at least one previously determined classification of signatures; creating a new classification with said signature if said closest proximity has a value less than a predetermined threshold value; and adding said signature to said one classification if said closest proximity has a value at least equal to said predetermined threshold value. - View Dependent Claims (9, 10, 11, 12, 13, 14)
-
-
15. Apparatus for performing pattern analysis of a digital signal containing at least one waveform, comprising:
-
means for detecting a fiducial point of each waveform of said digital signal, for capturing at least two characteristic features of each of said waveforms of said digital signal, and for creating a signature for each waveform based on said captured at least two characteristic features of each of said waveforms, and said means for detecting forms a delta value for each sampling point of said digital signal by subtracting the value of each sampling point from the value of the sampling point occurring m sampling points previous thereto, determining the average of n highest differences in values between consecuive sampling points in a predetermined interval of said digital signal, obtaining a trigger value based on said average, and detecting a first sampling point as being a fiducial point when a delta value is equal to or greater than said trigger value, where m and n are integers; means for determining a closest proximity of said signature to one classification of at least one previously determined classification of signatures; and means for storing said signature as a new classification if said closest proximity has a value less than a predetermined threshold value, and for storing said signature as part of an existing classification if said closest proximity has a value equal to or greater than said predetermined threshold value. - View Dependent Claims (16, 17)
-
-
18. Apparatus for performing pattern analysis of a digital signal containing at least one waveform, comprising:
-
means for detecting a fiducial point of each waveform of said digital signal, for capturing at least two characteristic features of each of said waveforms of said digital signal, and for creating a signature for each waveform based on said captured at least two characteristic features of each of said waveforms, and said means for detecting forms a delta value for each sampling point by subtracting the value of each sampling point from the value of the sampling point occurring m sampling points previous thereto, determining the average of n highest differences in values between consecutive sampling points in a predetermined interval of said digital signal, obtaining a trigger value based on said average, and determining a plurality of slope transitions and intervals in accordance with a relation between said delta values for the sampling points and said trigger value, where m and n are integers; means for determining a closest proximity of said signature to one classification of at least one previously determined classification of signatures; and means for storing said signature as a new classification if said closest proximity has a value less than a predetermined threshold value, and for storing said signature as part of an existing classification if said closest proximity has a value equal to or greater than said predetermined threshold value. - View Dependent Claims (19, 20, 21)
-
-
22. Apparatus for performing pattern analysis of a digital signal containing at least one waveform, comprising:
-
means for detecting a fiducial point of each waveform of said digital signal, for capturing at least two characteristic features of each of said waveforms of said digital signal including a plurality of interval features and a plurality of slope transition features separated by said interval features, for creating a signature for each waveform based on said captured at least two characteristic features of each of said waveforms, and for assigning a first value to each feature and a second value which differs from said first value by one to each said feature to provide a blurring effect to that minor deviation in patterns can be accounted for; means for determining a closest proximity of said signature to one classification of at least one previously determined classification of signatures; and means for storing said signature as a new classification if said closest proximity has a value less than a predetermined threshold value, and for storing said signature as part of an existing classification if said closest proximity has a value equal to or greater than said predetermined threshold value. - View Dependent Claims (23, 24)
-
Specification