# US 2015/0178631 A1 · Pattern recognition system

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2015/0178631 A1
- Document ID: US20150178631A1
- Status: Published application
- Kind: A1
- Application number: 14/472,247
- Filed: 2014-08-28
- Published: 2015-06-25
- Earliest claimed: 2013-09-04
- Assignees: Rokio, Inc.
- Applicants: Neural ID LLC
- Inventors: Tyson J. Thomas; Kristopher Robert Buschelman; Frank G. Evans; Karl P. Geiger; Michael P. Kelley; Eric C. Schneider; Timothy J. Carruthers; Jeffrey Brian Adams
- Web page: https://patentgrep.com/patent/US20150178631A1

## CPC

- G06N 20/00: Machine learning
- G06F 18/28: Determining representative reference patterns, e.g. by averaging or distorting; Generating dictionaries
- G06N 5/047: Pattern matching networks; Rete networks
- G06V 10/772: Determining representative reference patterns, e.g. averaging or distorting patterns; Generating dictionaries
- G06F 2218/10: by analysing the shape of a waveform, e.g. extracting parameters relating to peaks
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

Methods, apparatuses and systems directed to pattern learning, recognition, and metrology. In some particular implementations, the invention provides a flexible pattern recognition platform including pattern recognition engines that can be dynamically adjusted to implement specific pattern recognition configurations for individual pattern recognition applications. In certain implementations, the present invention provides for methods and systems suitable for analyzing and recognizing patterns in biological signals such as multi-electrode array waveform data. In other implementations, the present invention also provides for a partition configuration where knowledge elements can be grouped and pattern recognition operations can be individually configured and arranged to allow for multi-level pattern recognition schemes. In other implementations, the present invention provides methods and systems for dynamic learning of patterns in supervised and unsupervised manners.

## Claims

### Claim at position 1

1-102. (canceled)

### Claim 103

A pattern recognition system comprising:
one or more memories configured to maintain a knowledge element array and logic configured to:
create a pattern definition comprising a first set of one or more waveform characteristics representing a pattern of interest;
generate one or more knowledge elements with reference to the pattern definition for inclusion in the knowledge element array, each knowledge element comprising i) one or more operands collectively defining a data vector representing the pattern of interest and ii) a category identifier;
define a region of search relative to at least one of the waveform characteristics representing the pattern of interest, each occurrence of the region of search representing a data range in which to search for the pattern of interest;
perform a pattern recognition operation on input vectors derived from waveform data at least partially in an occurrence of the region of search by comparing the input vectors to the one or more knowledge elements of the knowledge element array.

### Claim 104 (depends on 103)

The pattern recognition system of claim 103, wherein the logic is further configured to define the region of search by: creating one or more marker definitions for one or more markers, wherein an occurrence of each of the one or more markers provides a reference for a corresponding occurrence of the region of search.

### Claim 105 (depends on 104)

The pattern recognition system of claim 104, wherein the one or more markers comprise a single marker, and wherein each occurrence of the single marker is associated with an occurrence of the region of search.

### Claim 106 (depends on 104)

The pattern recognition system of claim 104, wherein the range of waveform data represented by each occurrence of the region of search is defined by two consecutive occurrences of the single marker.

### Claim 107 (depends on 103)

The pattern recognition system of claim 103, wherein the logic is configured to prevent detection of an occurrence of the one or more markers in a user definable suppression interval.

### Claim 108 (depends on 103)

The pattern recognition system of claim 103, wherein the logic is configured to prevent detection of a new occurrence of the one or more markers with reference to a user definable hysteresis threshold.

### Claim 109 (depends on 103)

The pattern recognition system of claim 103, wherein the logic is configured to perform the pattern recognition operation by determining a match between each of the input vectors and the one or more knowledge elements of the knowledge element array if the input vector falls in any influence fields of the one or more knowledge elements.

### Claim 110 (depends on 103)

The pattern recognition system of claim 103, wherein the logic is further configured to generate a new knowledge element corresponding to an input vector, store the new knowledge element in the knowledge element array, and adjust one or more influence fields of other knowledge elements where the input vector is not recognized by the recognition operation.

### Claim 111 (depends on 103)

The pattern recognition system of claim 103, wherein the logic is further configured to adjust one or more influence fields of the knowledge elements, where the input vector does not represent the pattern of interest.

### Claim 112 (depends on 103)

The pattern recognition system of claim 103, wherein the logic is further configured to obtain one or more measurements from an occurrence of the pattern of interest recognized in the pattern recognition operation.

### Claim 113 (depends on 112)

The pattern recognition system of claim 112, wherein the logic is further configured to present the obtained measurements to a user.

### Claim 114 (depends on 103)

The pattern recognition system of claim 103, wherein the waveform characteristics comprise one or more of the following: peak amplitude derived from polynomial fit, sample count, pattern mean, median, standard deviation, minimum, maximum, slopes, pattern polynomial fit intercepts with baseline polynomial fit, maximum curvature elbows and widths (durations) at various relative amplitudes, QT interval, RR interval, RT interval, field potential duration, T wave peak amplitude, R wave peak amplitude, and total spike amplitude.

### Claim 115 (depends on 103)

The pattern recognition system of claim 103, wherein the logic is further configured to perform the pattern recognition operation by:
initially performing a macro-search that compares macro-search input vectors to both the first knowledge element and the one or more additional knowledge elements; and
if a first one of the macro-search input vectors matches one of the additional knowledge elements, performing a micro-search that compares micro-search input vectors to only the first knowledge element;
wherein
the micro-search input vectors are derived from a range of the test data near a range from which the first one of the macro-search input vector is derived, and
the micro-search uses a smaller step size than the macro-search.

### Claim 116 (depends on 115)

The pattern recognition system of claim 115, wherein the logic is further configured to perform the pattern recognition operation by recognizing an occurrence of the pattern of interest when the one of the micro-search input vector falls within an influence field of the one or more knowledge elements.

### Claim 117 (depends on 116)

The pattern recognition system of claim 116, wherein the logic is further configured to generate a new knowledge element corresponding to a first one of the micro-search input vectors into the knowledge element array as a knowledge element and adjust one or more influence fields of one or more other knowledge elements where the first micro-search input vector is not recognized by the pattern recognition operation.

### Claim 118

The pattern recognition system of 117, wherein the logic is further configured to adjust one or more influence fields of one or more of the knowledge elements where a first one of the micro-search input vectors does not represent the pattern of interest.

### Claim 119

A pattern recognition system comprising one or more memories configured to maintain a knowledge element array comprising a plurality of knowledge elements; and logic configured to:
determine one or more characteristics as features and/or properties of one or more patterns of interest from training examples;
select one or more channels of a multichannel system in which the determined characteristics are found;
perform a recognition or learning operation on data from the selected channels.

### Claim 120 (depends on 119)

The pattern recognition system of claim 119, wherein the training examples comprise data from a channel different from the selected one or more channels of the multichannel system.

### Claim 121 (depends on 119)

The pattern recognition system of claim 119, wherein the logic is further configured for processing data corresponding to different data types.

### Claim 122 (depends on 121)

The pattern recognition system of claim 121, wherein the logic is further configured to process data corresponding to data of one or more of the following types: biological, financial, marketing, machine control, industrial control system, software system, economic, epidemiological, audio, sonographic, electro-magnetic, radiographic, radiological, chromatographic, agricultural, astronomic, atmospheric, oceanographic, geologic, and geophysical data.

### Claim 123 (depends on 121)

The pattern recognition system of claim 121, wherein the different data types are derived from a plurality of different sensor types.

### Claim 124 (depends on 121)

The pattern recognition system of claim 121 wherein the plurality of different data types comprise any of image data, video data, audio data, waveform data, chemical data, text data, temperature data, or biological data.

### Claim 125 (depends on 119)

The pattern recognition system of any of claim 119, wherein the logic is configured to process input vectors corresponding to data of one or more of the following: electrocardiogram (ECG), electroencephalogram (EEG), multielectrode array, electromyogram (EMG), mechanomyogram (MMG), muscle tonus, muscle spasimus, electrooculograph (EOG), pressure and volume change (plethysmography), galvanic skin response, magnetoencephalogram (MEG), polymerase chain reaction, gas exchange, nutrient absorption, salt and chemical concentrations inside/outside cells, chemical exchange in media/reagents, ion exchange across ion channels, temperature, blood pressure, hydraulic pressure and water concentration, cell adhesion, cytometry, fluorescent signals, optogenetic signals, radionucleotide marker uptake/decay, functional and structural magnetic resonance imaging, and patterns of signals based on sensing diffusion in media (solid, liquid, gas, and plasma) and across osmotic membranes.

### Claim 126

A pattern recognition system comprising:
one or more memories configured to maintain a knowledge element array; and
logic configured to:
receive user input specifying a portion of training data having a pattern of interest;
extract one or more pattern characteristics from the portion of the training data;
create a pattern definition comprising the one or more pattern characteristics;
generate one or more knowledge elements with reference to the pattern definition for inclusion in the knowledge element array, each knowledge element comprising one or more operands collectively defining a data vector representing the pattern of interest; and
perform a pattern recognition operation on input vectors derived from test data by comparing the input vectors to the one or more knowledge elements of the knowledge element array.

### Claim 127

The pattern recognition system of 126, wherein the logic is further configured to present the training data using a graphical user interface.

### Claim 128

The pattern recognition system of 126, wherein the portion of training data having the pattern of interest is specified by the user using a graphical user interface.

### Claim 129 (depends on 126)

The pattern recognition system of claim 126, wherein, prior to performing the pattern recognition operation on the input vectors, the logic is further configured to:
create a processing template comprising a processing chain, and
apply the processing template by using at least part of the processing chain to process input data to derive the input vector.

### Claim 130 (depends on 129)

The pattern recognition system of claim 129, wherein the logic is further configured to store and reuse the processing template.

### Claim 131 (depends on 129)

The pattern recognition system of claim 129, wherein the processing template comprises a channel filter for filtering data from one or more channels of multichannel data.

### Claim 132 (depends on 129)

The pattern recognition system of claim 129, wherein the processing template comprises a global filter for filtering data from all channels of multichannel data including multielectrode array data.

### Claim 133 (depends on 129)

The pattern recognition system of claim 129, wherein the processing template comprises one or more feature extraction processes to extract features from the input data.
