Method and system for using a weighted response
First Claim
1. A method for analyzing an analyte, said method comprising:
- a) disposing a plurality of d sensors in an array, each sensor having a different response to said analyte;
b) exposing said d sensors to said analyte to generate a response vector for said analyte;
c) applying an importance index vector to said response vector to generate a weighted response vector; and
d) comparing said weighted response vector to a weighted training set and thereafter selecting a match to identify said analyte,wherein said importance index vector is generated after a training mode conducted on said d sensors and before identification of said analyte,wherein said importance index vector is the collection of importance indices for each sensor and the weighted training set is generated by application of the importance index to each said response vector for each exposure to each of said plurality of known analytes in a training set, andwherein said importance index for a particular sensor is calculated from all the responses for that sensor in a training set as follows;
4 Assignments
0 Petitions
Accused Products
Abstract
The present invention is related to the sensing of analytes and in particular, to methods of determining, resolving, identifying, or quantitating such analytes once detected. This invention provides systems and methods for identifying analytes, comprising using an importance index to give greater weight to the responses from sensors that measure characteristics most useful for identification of analytes. In other aspects, the systems and methods are useful for increasing the stability of electronic nose systems by prolonging the predictive capability of the training set of known analytes. In still other aspects, the systems and methods are useful for detecting and responding to events correlated with the presence of an analyte.
14 Citations
7 Claims
-
1. A method for analyzing an analyte, said method comprising:
-
a) disposing a plurality of d sensors in an array, each sensor having a different response to said analyte; b) exposing said d sensors to said analyte to generate a response vector for said analyte; c) applying an importance index vector to said response vector to generate a weighted response vector; and d) comparing said weighted response vector to a weighted training set and thereafter selecting a match to identify said analyte, wherein said importance index vector is generated after a training mode conducted on said d sensors and before identification of said analyte, wherein said importance index vector is the collection of importance indices for each sensor and the weighted training set is generated by application of the importance index to each said response vector for each exposure to each of said plurality of known analytes in a training set, and wherein said importance index for a particular sensor is calculated from all the responses for that sensor in a training set as follows; - View Dependent Claims (2)
-
-
3. A method for analyzing an analyte, said method comprising:
-
a) disposing a plurality of d sensors in an array, each sensor having a different response to said analyte; b) exposing said d sensors to said analyte to generate a response vector for said analyte; c) applying an importance index vector to said response vector to generate a weighted response vector; and d) comparing said weighted response vector to a weighted training set and thereafter selecting a match to identify said analyte, wherein said importance index vector is generated after a training mode conducted on said d sensors and before identification of said analyte, wherein said importance index vector is the collection of importance indices for each sensor and the weighted training set is generated by application of the importance index to each said response vector for each exposure to each of said plurality of known analytes in a training set, and wherein the application of said importance index vector to the response vector to generate a weighted response vector comprises; a) raising the importance index for each sensor to a certain power; b) multiplying each response in the response vector by the importance index for the corresponding sensor raised to a certain power to yield a weighted response; and c) assembling the weighted responses to generate a weighted response vector. - View Dependent Claims (5)
-
-
4. A method for analyzing an analyte, said method comprising:
-
a) disposing a plurality of d sensors in an array, each sensor having a different response to said analyte; b) exposing said d sensors to said analyte to generate a response vector for said analyte; c) applying an importance index vector to said response vector to generate a weighted response vector; and d) comparing said weighted response vector to a weighted training set and thereafter selecting a match to identify said analyte, wherein said importance index vector is generated after a training mode conducted on said d sensors and before identification of said analyte, wherein said importance index vector is the collection of importance indices for each sensor and the weighted training set is generated by application of the importance index to each said response vector for each exposure to each of said plurality of known analytes in a training set, and wherein the application of said importance index vector to the training set to generate a weighted training set comprises; a) raising the importance index for each sensor to a certain power; b) multiplying each response in the response vector for the training set by said importance index for the corresponding sensor raised to a certain power to yield a weighted response; and c) assembling the weighted responses to generate a weighted response vector.
-
-
6. A method for identifying an unknown analyte, comprising:
-
a) disposing a plurality of d sensors in a array, each having a different response to different analytes; b) exposing said d sensors two or more times to a plurality of known analytes and measuring the responses from each sensor to each exposure to create a first training set, wherein said first training set contains a response vector for each exposure to each of said plurality of known analytes; c) generating a first importance index for each sensor; d) selecting a subset of sensors having first importance indices above a certain threshold; e) generating a second training set, wherein said second training set consists of all the response vectors in the first training set but with responses to those sensors not passing the selection criterion removed; f) calculating a second importance index based on the second training set and assembling a second importance index vector; g) applying the second importance index vector to the second training set to generate a second weighted training set; h) exposing said subset of sensors to said unknown analyte to generate a response vector for said unknown analyte; i) applying said second importance index vector to said response vector for the unknown analyte to obtain a weighted response vector; and j) comparing said weighted response vector to the second weighted training set, and then selecting the best match thereby identifying said analyte.
-
-
7. A method for identifying an unknown analyte, said method comprising:
-
a) disposing a plurality of d sensors in a array, each having a different response to different analytes; b) exposing said d sensors two or more times to a plurality of known analytes and measuring the responses from each sensor to each exposure to create a first training set, wherein said first training set contains a response vector for each exposure to each of said plurality of known analytes; c) generating an first importance index for each sensor; d) identifying a subset of sensors with importance indices above a certain threshold; e) generating a second training set, wherein said second training set consists of all the response vectors in the first training set but with responses to those sensors not passing the selection criterion removed; f) exposing said subset of sensors to said unknown analyte to generate a response vector; and g) comparing said response vector for the unknown analyte to response vectors in the second training set, and thereafter selecting the best match to identify said unknown analyte.
-
Specification