CN107220670B - Method for extracting characteristics of supervised artificial taste system based on discrete wavelet transform - Google Patents

Method for extracting characteristics of supervised artificial taste system based on discrete wavelet transform Download PDF

Info

Publication number
CN107220670B
CN107220670B CN201710391480.2A CN201710391480A CN107220670B CN 107220670 B CN107220670 B CN 107220670B CN 201710391480 A CN201710391480 A CN 201710391480A CN 107220670 B CN107220670 B CN 107220670B
Authority
CN
China
Prior art keywords
wavelet
calculating
artificial taste
coefficient
discrimination
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201710391480.2A
Other languages
Chinese (zh)
Other versions
CN107220670A (en
Inventor
刘涛
陈艳兵
武萌雅
李东琦
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Chongqing University
Original Assignee
Chongqing University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Chongqing University filed Critical Chongqing University
Priority to CN201710391480.2A priority Critical patent/CN107220670B/en
Publication of CN107220670A publication Critical patent/CN107220670A/en
Application granted granted Critical
Publication of CN107220670B publication Critical patent/CN107220670B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G06F18/2155—Generating training patterns; Bootstrap methods, e.g. bagging or boosting characterised by the incorporation of unlabelled data, e.g. multiple instance learning [MIL], semi-supervised techniques using expectation-maximisation [EM] or naïve labelling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques

Abstract

The invention belongs to the technical field of gas phase chemical analysis, and discloses a method for extracting characteristics of a supervised artificial taste system based on discrete wavelet transform, which comprises the following steps: regarding any working electrode in the electrode array of the artificial taste sense system, regarding the time domain original response of the working electrode as a time sequence and performing discrete wavelet change to obtain a series of wavelet coefficients, and converting the wavelet coefficients into energy ratios; working electrodes are arranged in the electrode array of the artificial taste sense system, and the t-th sample mean value and all sample mean values are calculated; calculating the average distance in the classes and the average distance between the classes; solving a discrimination coefficient; and calculating the total discrimination coefficient and the discrimination coefficients corresponding to all the wavelet coefficients to obtain the optimal energy ratio. The invention reduces the data complexity, and reduces the original data dimension from 228000 (30000 x 6+1500 x 32) to 6; in addition, the identification accuracy of the KNN method is greatly improved.

Description

Method for extracting characteristics of supervised artificial taste system based on discrete wavelet transform
Technical Field
The invention belongs to the technical field of gas phase chemical analysis, and particularly relates to a method for extracting characteristics of a supervised artificial taste system based on discrete wavelet transform.
Background
The artificial gustatory system is a novel gas phase chemical analysis means, and has the advantages of rapid detection, no wound, simple and convenient operation, low cost and the like. The artificial gustation system is mainly divided into two parts of an electrode array and a mode recognition method, wherein the electrode array mostly adopts an inert metal electrode with cross sensitivity to form a three-electrode system, and the electrodes in the three-electrode system can be divided into the following parts according to the difference of self functions: a reference electrode, an auxiliary electrode and a working electrode. The system forms an electric field in a solution through a reference electrode, an auxiliary electrode and a working electrode form a loop under the action of the electric field, and the mode recognition method distinguishes and identifies substances according to different loop current responses. Because electric fields with different frequencies and intensities can generate rich transient response of working electrode current, the current artificial gustation system mostly adopts 'multi-frequency pulse' to carry out electric field modulation on a reference electrode, response with rich frequency components can be obtained on the working electrode, and an original signal is usually represented in a time domain form and needs to be subjected to feature extraction through a frequency domain analysis method such as discrete wavelet transformation so as to be convenient for distinguishing a 'pattern recognition' part. The traditional discrete wavelet transform can generate a large number of wavelet coefficients along with the increase of the number of decomposition layers, the complexity of 'pattern recognition' can be increased by selecting excessive wavelet coefficients as features, and the possibility of missing the selection of the optimal wavelet coefficients exists due to the fact that the selection amount is too small.
In summary, the problems of the prior art are as follows: in the traditional discrete wavelet transform, too many wavelet coefficients are selected as features, so that the complexity of pattern recognition is increased, and the selection quantity is too small and the optimal wavelet coefficients are missed.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a method for extracting the characteristics of a supervised artificial taste system based on discrete wavelet transform.
The invention is realized in this way, a method for extracting the characteristics of the supervised artificial gustation system based on discrete wavelet transform comprises: regarding any working electrode in the electrode array of the artificial taste sense system, regarding the time domain original response of the working electrode as a time sequence and performing discrete wavelet change to obtain a series of wavelet coefficients, and converting the wavelet coefficients into energy ratios; working electrodes are arranged in the electrode array of the artificial taste sense system, and the t-th sample mean value and all sample mean values are calculated; calculating the average distance in the classes and the average distance between the classes; solving a discrimination coefficient; and calculating the total discrimination coefficient and the discrimination coefficients corresponding to all the wavelet coefficients to obtain the optimal energy ratio.
Further, the method for extracting the characteristics of the supervised artificial taste system based on the discrete wavelet transform comprises a wavelet coefficient conversion stage and an optimal energy ratio searching stage.
Further, the wavelet coefficient conversion stage comprises:
obtaining wavelet coefficients
Figure BDA0001307621210000021
Wherein i is the number of discrete wavelet transform decomposition layers, j is the jth wavelet coefficient after the ith layer decomposition, j ∈ N is less than or equal to 2i(ii) a Wavelet coefficient
Figure BDA0001307621210000022
Conversion to energy ratio as follows
Figure BDA0001307621210000023
Figure BDA0001307621210000024
Further, the optimal energy ratio searching stage comprises:
1) order to
Figure BDA0001307621210000025
Calculating the t-th sample mean value
Figure BDA0001307621210000026
And all sample means
Figure BDA0001307621210000027
Figure BDA0001307621210000028
Figure BDA0001307621210000029
Wherein
Figure BDA00013076212100000210
Characteristic value, M, representing the mth training sampletEach represents the number of the t-th type samples;
2) calculating the average distance d in classwAnd the average distance d between classesb：
Figure BDA00013076212100000211
Figure BDA0001307621210000031
3) Calculating the discrimination coefficient of the jth wavelet coefficient of the ith layer to the tth sample
Figure BDA0001307621210000032
Figure BDA0001307621210000033
4) Repeating the steps 1) to 3), and solving the corresponding T-type samples
Figure BDA0001307621210000034
Calculating the total discrimination coefficient theta of the jth wavelet coefficient of the ith layeri,j：
Figure BDA0001307621210000035
5) Repeating the steps 1) to 4), and calculating the total discrimination coefficient theta corresponding to all the wavelet coefficientsi,j；
6) Finding the optimal energy ratio*：
(i*,j*)＝arg min θi,j；
Figure BDA0001307621210000036
Wherein (i)*,j*) The wavelet coefficient representing the minimum total discrimination coefficient is the ith*Layer j (j)*The number of the main components is one,*and the final characteristic extraction result is used as the input of a subsequent pattern recognition method.
Another objective of the present invention is to provide an artificial taste system applying the supervised artificial taste system feature extraction method for discrete wavelet transform.
The invention has the advantages and positive effects that: the data complexity is reduced, the original data dimension is reduced from 228000 (30000 x 6+1500 x 32) to 6, and the complexity and the calculation cost of a subsequent pattern recognition method are reduced; when a nearest neighbor (KNN) method is adopted as a subsequent pattern recognition method, the average recognition accuracy of the multi-class sample recognition system can be increased from 57.14% to 83.33%.
Drawings
FIG. 1 is a flow chart of a method for extracting features of a supervised artificial taste system based on discrete wavelet transform according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail with reference to the following embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The following detailed description of the principles of the invention is provided in connection with the accompanying drawings.
As shown in fig. 1, the method for extracting features of a supervised artificial gustatory system based on discrete wavelet transform according to the embodiment of the present invention includes the following steps:
s101: regarding any working electrode in the electrode array of the artificial taste sense system, regarding the time domain original response of the working electrode as a time sequence and performing discrete wavelet change to obtain a series of wavelet coefficients, and converting the wavelet coefficients into energy ratios;
s102: working electrodes are arranged in the electrode array of the artificial taste sense system, and the t-th sample mean value and all sample mean values are calculated; calculating the average distance in the classes and the average distance between the classes; solving a discrimination coefficient; and calculating the total discrimination coefficient and the discrimination coefficients corresponding to all the wavelet coefficients to obtain the optimal energy ratio.
The method for extracting the characteristics of the supervised artificial taste system based on the discrete wavelet transform provided by the embodiment of the invention is divided into two stages of wavelet coefficient conversion and optimal energy ratio search.
1. Wavelet coefficient conversion stage
For artificial flavorAny working electrode k in the electrode array of the sensing system takes the time domain original response as a time sequence and carries out discrete wavelet change to obtain a series of wavelet coefficients
Figure BDA0001307621210000041
Wherein i is the number of discrete wavelet transform decomposition layers, j is the jth wavelet coefficient after the ith layer decomposition, j ∈ N is less than or equal to 2i. Wavelet coefficient
Figure BDA0001307621210000042
Conversion to energy ratio as follows
Figure BDA0001307621210000043
Figure BDA0001307621210000044
2. Optimal energy ratio search phase
Assuming that there are K working electrodes in the electrode array of the artificial taste system
Figure BDA0001307621210000045
There are M training samples of the T class, wherein the T class sample MtAnd then:
1) calculating the t-th sample mean value and all sample mean values:
Figure BDA0001307621210000051
Figure BDA0001307621210000052
wherein
Figure BDA0001307621210000053
Representing the mth training sample.
2) Calculating the average distance d in classwAnd the average distance d between classesb：
Figure BDA0001307621210000054
Figure BDA0001307621210000055
3) Calculating the degree of discrimination
Figure BDA0001307621210000056
Figure BDA0001307621210000057
4) Repeating the steps 1) to 3), and solving all T
Figure BDA0001307621210000058
Calculating the total discrimination coefficient thetai,j：
Figure BDA0001307621210000059
5) Repeating the steps 1) to 4), and calculating the discrimination coefficient theta corresponding to all the wavelet coefficientsi,j。
6) Finding the optimal energy ratio*：
(i*,j*)＝arg min θi,j；
Figure BDA00013076212100000510
Will eventually be*As a result of the feature extraction, as input for a subsequent "pattern recognition" method.
The application of the principles of the present invention will now be described in further detail with reference to specific embodiments.
The embodiment of the invention uses an artificial olfaction system to analyze 7 substances such as black tea, green tea, Pu' er tea, oolong tea, red wine, white spirit, beer and the like. Data were collected 9 times for each material at different concentrations. In each data acquisition process, the number of the working electrodes is 6, the sampling rate of the system is 20Hz, the acquisition duration is 25min, 30000 data are acquired by each electrode in total, and the excitation pulse length of the electrode array is 150. In order to verify the effect of the involved data fusion method, 6 times of data acquisition results are randomly selected from each substance for training, the rest 3 times are used as test samples, and a nearest neighbor algorithm KNN is used as a classifier for identification. The number of wavelet decomposition layers was 10, and 4 experiments were performed. For comparison, two feature extraction methods were used in table 1: "method 1" is a proposed method, and method 2 is a method of directly using the original data as KNN input data, and as can be seen from the results in the table, the proposed method reduces the complexity of data, and reduces the original data dimension from 228000 (30000 × 6+1500 × 32) to 6; in addition, the identification accuracy of the KNN method is greatly improved.
TABLE 1 identification accuracy comparison Table
Figure BDA0001307621210000061
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.

Claims (3)

1. A method for extracting features of a supervised artificial taste system based on discrete wavelet transform is characterized by comprising the following steps: regarding any working electrode in the electrode array of the artificial taste sense system, regarding the time domain original response of the working electrode as a time sequence and performing discrete wavelet change to obtain a series of wavelet coefficients, and converting the wavelet coefficients into energy ratios; working electrodes are arranged in the electrode array of the artificial taste sense system, and the t-th sample mean value and all sample mean values are calculated; calculating the average distance in the classes and the average distance between the classes; solving a discrimination coefficient; calculating the total discrimination coefficient and the discrimination coefficients corresponding to all wavelet coefficients to obtain the optimal energy ratio;
the method for extracting the characteristics of the supervised artificial taste system based on the discrete wavelet transform comprises a wavelet coefficient conversion stage and an optimal energy ratio searching stage;
the optimal energy ratio searching stage comprises:
1) calculating the t-th sample mean value and all sample mean values:
Figure FDA0002431763820000011
Figure FDA0002431763820000012
wherein
Figure FDA0002431763820000013
Represents the mth training sample;
2) calculating the average distance d in classwAnd the average distance d between classesb：
Figure FDA0002431763820000014
Figure FDA0002431763820000015
3) Calculating the degree of discrimination
Figure FDA0002431763820000016
Figure FDA0002431763820000021
4) Repeating the steps 1) to 3), and solving all T
Figure FDA0002431763820000022
Calculating the total discrimination coefficient thetai,j：
Figure FDA0002431763820000023
5) Repeating the steps 1) to 4), and calculating the discrimination coefficient theta corresponding to all the wavelet coefficientsi,j；
6) Finding the optimal energy ratio*：
(i*,j*)＝argminθi,j；
Figure FDA0002431763820000024
Will be provided with*As a result of the feature extraction, as an input for a subsequent pattern recognition method.
2. The method of discrete wavelet transform-based supervised artificial taste system feature extraction as recited in claim 1, wherein the wavelet coefficient transformation stage comprises:
obtaining wavelet coefficients
Figure FDA0002431763820000025
Wherein i is the number of discrete wavelet transform decomposition layers, j is the jth wavelet coefficient after the ith layer decomposition, j ∈ N is less than or equal to 2i(ii) a Wavelet coefficient
Figure FDA0002431763820000026
Converted to energy ratio according to
Figure FDA0002431763820000027
Figure FDA0002431763820000028
3. An artificial taste system using the supervised artificial taste system feature extraction method of discrete wavelet transform as claimed in any one of claims 1 to 2.
CN201710391480.2A 2017-05-27 2017-05-27 Method for extracting characteristics of supervised artificial taste system based on discrete wavelet transform Active CN107220670B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710391480.2A CN107220670B (en) 2017-05-27 2017-05-27 Method for extracting characteristics of supervised artificial taste system based on discrete wavelet transform

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710391480.2A CN107220670B (en) 2017-05-27 2017-05-27 Method for extracting characteristics of supervised artificial taste system based on discrete wavelet transform

Publications (2)

Publication Number Publication Date
CN107220670A CN107220670A (en) 2017-09-29
CN107220670B true CN107220670B (en) 2020-07-14

Family

ID=59948422

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710391480.2A Active CN107220670B (en) 2017-05-27 2017-05-27 Method for extracting characteristics of supervised artificial taste system based on discrete wavelet transform

Country Status (1)

Country Link
CN (1) CN107220670B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109239585A (en) * 2018-09-06 2019-01-18 南京理工大学 A kind of method for diagnosing faults based on the preferred wavelet packet of improvement
CN112910813B (en) * 2021-04-10 2022-09-06 青岛科技大学 LDA-KNN-based underwater sound signal automatic modulation identification method

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101692053A (en) * 2009-10-09 2010-04-07 江苏大学 Multi-sensing information fusion based instrumental intelligent evaluation method for quality of famous tea
CN102520341A (en) * 2011-12-05 2012-06-27 南京航空航天大学 Analog circuit fault diagnosis method based on Bayes-KFCM (Kernelized Fuzzy C-Means) algorithm
CN103234918A (en) * 2013-04-12 2013-08-07 泸州品创科技有限公司 Method for discriminating liquor odor by liquid array taste sense bionic sensor
CN104268575A (en) * 2014-09-29 2015-01-07 西南大学 Electronic nose signal processing method based on mixing characteristic matrix
US9201068B2 (en) * 2008-09-03 2015-12-01 Clarkson University Bioelectronic tongue for food allergy detection
CN105160359A (en) * 2015-09-17 2015-12-16 华南理工大学 Complex structure damage cooperative identification method based on ultrasonic guided-wave
CN106096649A (en) * 2016-06-08 2016-11-09 北京科技大学 Sense of taste induced signal otherness feature extracting method based on core linear discriminant analysis
CN106093120A (en) * 2016-05-27 2016-11-09 北京科技大学 A kind of Feature Parameter Fusion method of Electronic Nose response signal in tea grades identification
CN106127259A (en) * 2016-07-04 2016-11-16 西南大学 E-nose signal Feature fusion based on degree of isolation and distinctiveness ratio

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9201068B2 (en) * 2008-09-03 2015-12-01 Clarkson University Bioelectronic tongue for food allergy detection
CN101692053A (en) * 2009-10-09 2010-04-07 江苏大学 Multi-sensing information fusion based instrumental intelligent evaluation method for quality of famous tea
CN102520341A (en) * 2011-12-05 2012-06-27 南京航空航天大学 Analog circuit fault diagnosis method based on Bayes-KFCM (Kernelized Fuzzy C-Means) algorithm
CN103234918A (en) * 2013-04-12 2013-08-07 泸州品创科技有限公司 Method for discriminating liquor odor by liquid array taste sense bionic sensor
CN104268575A (en) * 2014-09-29 2015-01-07 西南大学 Electronic nose signal processing method based on mixing characteristic matrix
CN105160359A (en) * 2015-09-17 2015-12-16 华南理工大学 Complex structure damage cooperative identification method based on ultrasonic guided-wave
CN106093120A (en) * 2016-05-27 2016-11-09 北京科技大学 A kind of Feature Parameter Fusion method of Electronic Nose response signal in tea grades identification
CN106096649A (en) * 2016-06-08 2016-11-09 北京科技大学 Sense of taste induced signal otherness feature extracting method based on core linear discriminant analysis
CN106127259A (en) * 2016-07-04 2016-11-16 西南大学 E-nose signal Feature fusion based on degree of isolation and distinctiveness ratio

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
"A feature extraction method based on wavelet packet analysis for discrimination of Chinese vinegars using a gas sensors array";Yong Yin, Huichun Yu, Hongshun Zhang;《Sensors and Actuators B: Chemical》;20080731;第1005-1009页 *

Also Published As

Publication number Publication date
CN107220670A (en) 2017-09-29

Similar Documents

Publication Publication Date Title
CN108718288B (en) Method for recognizing digital signal modulation mode based on convolutional neural network
CN110197209B (en) Radiation source identification method based on multi-feature fusion
CN101919695A (en) Electrocardiosignal QRS complex detection method based on wavelet transform
CN107220670B (en) Method for extracting characteristics of supervised artificial taste system based on discrete wavelet transform
Hati et al. Plant recognition from leaf image through artificial neural network
CN103902987A (en) Station caption identifying method based on convolutional network
CN110084812B (en) Terahertz image defect detection method, device and system and storage medium
CN107798351B (en) Deep learning neural network-based identity recognition method and system
CN106357575A (en) Multi-parameter jointly-estimated interference type identification method
CN113283331A (en) Multi-class target identification method and system for unattended sensor system
CN106295708B (en) A kind of continuous data preprocess method based on Fisher classifier group
CN116127288B (en) Nanopore sensing signal noise removing method and device based on independent component analysis
CN107528646B (en) Interference signal identification and extraction method based on broadband spectrum
CN104125050B (en) Ultrahigh frequency RFID card reader protocol conformance test method
CN114186596B (en) Multi-window identification method and device for spectrogram peaks and electronic equipment
CN114239657A (en) Time sequence signal identification method based on complex value interference neural network
CN108507607B (en) Weak signal detection method based on kernel function
CN111464468A (en) Signal modulation mode identification method, device, equipment and readable storage medium
CN107894837A (en) Dynamic sentiment analysis model sample processing method and processing device
CN109598216B (en) Convolution-based radio frequency fingerprint feature extraction method
CN108008262B (en) Rapid partial discharge detection method
CN110336638A (en) A kind of short-term burst signal detecting method based on time-frequency figure
CN115100733A (en) RFID gesture recognition method, computer device, product and storage medium
CN114358041A (en) Characteristic waveform extraction method and analysis method based on hybrid algorithm
CN104820074A (en) Electronic nose characteristic extraction method based on moving window function

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant