CN104820074B - Electronic nose characteristic extraction method based on moving window function - Google Patents
Electronic nose characteristic extraction method based on moving window function Download PDFInfo
- Publication number
- CN104820074B CN104820074B CN201510252261.7A CN201510252261A CN104820074B CN 104820074 B CN104820074 B CN 104820074B CN 201510252261 A CN201510252261 A CN 201510252261A CN 104820074 B CN104820074 B CN 104820074B
- Authority
- CN
- China
- Prior art keywords
- electronic nose
- window function
- window
- characteristic
- response curve
- 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.)
- Expired - Fee Related
Links
Landscapes
- Image Processing (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides an electronic nose characteristic extraction method based on a moving window function. The method comprises the following steps: placing a window function at an initial preset position of an electronic nose response curve, and taking the area encircled by the window function and the electronic nose response curve as a characteristic component; moving the window function along a time axis, and acquiring a characteristic vector constructed by multiple characteristic components; setting a weight coefficient for the characteristic component of each sensor, introducing a quantum particle swarm algorithm, and optimizing the weight coefficient of the characteristic vectors of multiple sensors during weighted handling; and performing intelligent identification on the weighted characteristic vectors by utilizing a support vector machine algorithm. According to the method, the technical problems that the time domain information is lost due to variation domain characteristics and the steady state response and transient state response of the whole response curve cannot be reflected are solved, different electronic nose signals are distinguished by responding different window functions by utilizing an electronic nose array signal, the identification rate of the electronic nose is increased by introducing the quantum particle swarm algorithm and the support vector machine algorithm, and the identification effect is good.
Description
Technical field
The present invention relates to e-nose signal and field of information processing, and in particular to a kind of Electronic Nose based on mobile window function
Feature extracting method.
Background technology
Electronic Nose reliably and can be realized quickly to simple as a kind of intelligent apparatus of mimic biology olfactory system
Or complicated smell is distinguished, relative to the expensive gas analysis instrument such as traditional gas chromatograph, it is simple to operate, analysis
Reliable results, and it is adapted to Site Detection, thus it is widely used in the fields such as food, agricultural, medical treatment, environment measuring.Feature is carried
It is two key components in an electric nasus system to take with pattern-recognition, and wherein feature extraction largely have impact on classification
The reliability of model and the accuracy to unknown sample identification, pattern-recognition is that the information after feature extraction is carried out again suitably
Process, so as to obtain accurate gas componant and concentration information.
Existing e-nose signal feature extracting method mainly has three classes：1st, based on original response curve, the method is directed to gas
The original response of body sensor carries out feature extraction, and common feature includes：Maximum (steady-state response), the rate of rise, response are bent
Line integral etc.；2nd, based on curve matching, the method carries out curve fitting first to original response curve, is then made with fitting coefficient
It is characterized, conventional model of fit has fitting of a polynomial, exponential function fitting, Fraction Functions fitting, arctan function fitting etc.；
3rd, the feature extraction based on transform domain, the method carries out first certain specific conversion to sensor array original response, then
The later coefficient of conversion is extracted as feature, conventional transform domain feature includes：Fourier Transform Coefficients, wavelet conversion coefficient
Deng.
Weak point：The feature of extraction can not reflect the steady-state response of whole response curve and transient response information, and
Feature extraction based on transform domain can lose time-domain information, it is impossible to the feature of the complete whole response process of reflection；And it is only sharp
Extracting feature with Electronic Nose response signal itself can not reflect response situation of the Electronic Nose array signal to other signals.
The content of the invention
The application must not reflect whole by providing a kind of Electronic Nose feature extracting method based on mobile window function with solution
Individual response curve steady-state response and transient response information, domain of variation Character losing time-domain information, it is impossible to which complete reflection is whole to be responded
Process feature, and technical problem of the Electronic Nose array signal to the response situation of other signals can not be reflected.
To solve above-mentioned technical problem, the application is employed the following technical solutions and is achieved：
A kind of Electronic Nose feature extracting method based on mobile window function, comprises the following steps：
S1：Select a window function and place it on the initial precalculated position of Electronic Nose response curve, with window function
The area surrounded with Electronic Nose response curve is used as characteristic component；
S2：In units of default time span, window function to the left or/and is moved right successively along time shaft, and will
The multiple characteristic components for obtaining are built into characteristic vector；
S3：The characteristic vector of each sensor is set into a weight coefficient, quanta particle swarm optimization is introduced, will be multiple
The weight coefficient when characteristic vector of sensor is weighted process is optimized；
S4：Intelligent Recognition is carried out to the characteristic vector after weighting using algorithm of support vector machine.
Window function in said method extracts the time-domain information of original response curve similar to a wave filter, rather than
Frequency domain information.Mobile window function catches feature extracting method and solves domain of variation Character losing time-domain information, it is impossible to which reflection is whole
The problem of response curve steady-state response and transient response.Simultaneously using quantum particle swarm to the vectorial weight coefficient of sensor characteristics
It is optimized, Intelligent Recognition is carried out to the characteristic vector after weighting using algorithm of support vector machine, so that discrimination is significantly
Improve.
Further, the concrete grammar of step S3 is to give a weight coefficient ω to each sensor i firsti,
N weight coefficient of n sensor constitutes a weighing vector W=[ω1,ω2,…ωn], wherein ωiFor positive count, so
Afterwards weighting is processed into later sensor characteristics input grader identification, i.e., find each sensor with quanta particle swarm optimization
ωi, so as to find so that the W when fitness function of quanta particle swarm optimization reaches maximum, fitness function herein is
Grader discrimination.
Used as a kind of preferred technical scheme, initial precalculated position is at the peak value of Electronic Nose response curve in step S1.
Because including more key message at response curve peak value.
Used as a kind of preferred technical scheme, default time span is the window width of window function in step S2.
Used as a kind of preferred technical scheme, the nuclear parameter and punishment parameter in step S4 in SVMs passes through quantum
Population is optimized.
Window function can be the window of any point number width, it is possible to be placed on the optional position of response curve, further,
Window function can be quarter window, Blackman window, hamming code window, Hanning window, rectangular window or Gaussian window.By changing window function class
Type, width, placement location, the recognition effect of Electronic Nose is different.Overcome and extract spy in itself just with Electronic Nose response signal
Levy the technical problem that can not reflect Electronic Nose array signal to the response situation of other signals.
Compared with prior art, the technical scheme that the application is provided, the technique effect or advantage having be：This method is solved
Domain of variation Character losing time-domain information, it is impossible to reflect the technical problem of whole response curve steady-state response and transient response, and
Different e-nose signals are distinguished the reaction between different window functions using Electronic Nose array signal, while introducing quantum grain
Swarm optimization and algorithm of support vector machine improve the discrimination of Electronic Nose, and recognition effect is better than other feature extracting methods.
Description of the drawings
Fig. 1 is the workflow diagram of the present invention；
Fig. 2 is the window function trapping schematic diagram of the present invention；
Fig. 3 is the mobile window function trapping schematic diagram of the present invention.
Specific embodiment
The embodiment of the present application, must not with solution by providing a kind of Electronic Nose feature extracting method based on mobile window function
Reflect whole response curve steady-state response and transient response information, domain of variation Character losing time-domain information, it is impossible to completely reflect whole
Individual response process feature, and technical problem of the Electronic Nose array signal to the response situation of other signals can not be reflected.
It is right below in conjunction with Figure of description and specific embodiment in order to be better understood from above-mentioned technical proposal
Above-mentioned technical proposal is described in detail.
Embodiment
A kind of Electronic Nose feature extracting method based on mobile window function, as shown in figure 1, comprising the following steps：
S1：Select a window function and place it on the initial precalculated position of Electronic Nose response curve, with window function
The area surrounded with Electronic Nose response curve is used as characteristic component；
S2：In units of default time span, window function to the left or/and is moved right successively along time shaft, and will
The multiple characteristic components for obtaining are built into characteristic vector；
S3：The characteristic vector of each sensor is set into a weight coefficient, quanta particle swarm optimization is introduced, will be multiple
The weight coefficient when characteristic vector of sensor is weighted process is optimized；
S4：Intelligent Recognition is carried out to the characteristic vector after weighting using algorithm of support vector machine.
Window function in said method extracts the time-domain information of original response curve similar to a wave filter, rather than
Frequency domain information.Mobile window function catches feature extracting method and solves domain of variation Character losing time-domain information, it is impossible to which reflection is whole
The problem of response curve steady-state response and transient response.Simultaneously using quantum particle swarm to the vectorial weight coefficient of sensor characteristics
It is optimized, Intelligent Recognition is carried out to the characteristic vector after weighting using algorithm of support vector machine, so that discrimination is significantly
Improve.
Further, the concrete grammar of step S3 is to give a weight coefficient ω to each sensor i firsti,
N weight coefficient of n sensor constitutes a weighing vector W=[ω1,ω2,…ωn], wherein ωiFor positive count, so
Afterwards weighting is processed into later sensor characteristics input grader identification, i.e., find each sensor with quanta particle swarm optimization
ωi, so as to find so that the W when fitness function of quanta particle swarm optimization reaches maximum, fitness function herein is
Grader discrimination.
Used as a kind of preferred technical scheme, initial precalculated position is at the peak value of Electronic Nose response curve in step S1.
Because including more key message at response curve peak value.As shown in Fig. 2 being placed on Electronic Nose response curve for window function
Peak value at seizure schematic diagram.
As a kind of preferred technical scheme, in step S2 default time span for window function window width, such as Fig. 3
It is shown.
Used as a kind of preferred technical scheme, the nuclear parameter and punishment parameter in step S4 in SVMs passes through quantum
Population is optimized.
Window function can be the window of any point number width, it is possible to be placed on the optional position of response curve, further,
Window function can be quarter window, Blackman window, hamming code window, Hanning window, rectangular window or Gaussian window.By changing window function class
Type, width, placement location, the recognition effect of Electronic Nose is different.Overcome and extract spy in itself just with Electronic Nose response signal
Levy the technical problem that can not reflect Electronic Nose array signal to the response situation of other signals.
In order to prove to move the high efficiency of window function trapping, we will move window function trapping and other feature extractions
And the method for incorporating quantum particle group optimizing and support vector machine classifier is compared, such as peak value, the rate of rise, lower drop angle
Rate, Fourier transform, wavelet transformation and window function catch.From table 1 it follows that the recognition effect of mobile window function trapping
The method of other feature extractions is better than, wherein peak value and the rise time realizes that discrimination is 87.5%, fall time obtains knowledge
Rate is not 85.5%, and Fourier transformation and wavelet transformation discrimination are 90.0%, and window function catches and mobile window function catches
Discrimination be respectively 95.0% and 97.5%.
The different characteristic extracting method discrimination contrast table of table 1
In above-described embodiment of the application, there is provided a kind of Electronic Nose feature extracting method based on mobile window function, will
Window function is placed on the initial precalculated position of Electronic Nose response curve, with the area that window function and Electronic Nose response curve are surrounded
As characteristic component；Window function is moved along time shaft, multiple characteristic components is obtained and is built into characteristic vector；By each sensor
Characteristic vector sets a weight coefficient, introduces quanta particle swarm optimization, and the characteristic vector of multiple sensors is weighted
Weight coefficient during process is optimized；Intelligent Recognition is carried out to the characteristic vector after weighting using algorithm of support vector machine.This
Method solves domain of variation Character losing time-domain information, it is impossible to reflect the technology of whole response curve steady-state response and transient response
Problem, and using Electronic Nose array signal reacting to distinguish different e-nose signals between different window functions, while introducing
Quanta particle swarm optimization and algorithm of support vector machine improve the discrimination of Electronic Nose, and recognition effect is good.
It should be pointed out that described above is not limitation of the present invention, the present invention is also not limited to the example above,
Change, modified, addition or replacement that those skilled in the art are made in the essential scope of the present invention, also should
Belong to protection scope of the present invention.
Claims (6)
1. it is a kind of based on the Electronic Nose feature extracting method for moving window function, it is characterised in that to comprise the following steps：
S1：Select a window function and place it on the initial precalculated position of Electronic Nose response curve, with window function and electricity
The area that sub- nose response curve is surrounded is used as characteristic component；
S2：In units of default time span, window function to the left or/and is moved right successively along time shaft, and will be obtained
Multiple characteristic components be built into characteristic vector；
S3：The characteristic vector of each sensor is set into a weight coefficient, quanta particle swarm optimization is introduced, by multiple sensings
The weight coefficient when characteristic vector of device is weighted process is optimized；
S4：Intelligent Recognition is carried out to the characteristic vector after weighting using algorithm of support vector machine.
2. it is according to claim 1 based on the Electronic Nose feature extracting method for moving window function, it is characterised in that the step
The concrete grammar of rapid S3 is to give a weight coefficient ω to each sensor i firsti, n weight coefficient of n sensor
Constitute a weighing vector W=[ω1,ω2,…ωn], wherein ωiFor positive count, then weighting is processed into later sensing
The input grader identification of device feature, i.e., find the ω of each sensor with quanta particle swarm optimizationi, so as to find so that quantum grain
The fitness function of swarm optimization reaches W when maximum, and fitness function herein is grader discrimination.
3. it is according to claim 1 based on the Electronic Nose feature extracting method for moving window function, it is characterised in that step S1
In initial precalculated position for the peak value of Electronic Nose response curve at.
4. it is according to claim 1 based on the Electronic Nose feature extracting method for moving window function, it is characterised in that step S2
In default time span for window function window width.
5. it is according to claim 1 based on the Electronic Nose feature extracting method for moving window function, it is characterised in that step S4
Nuclear parameter and punishment parameter in middle SVMs is optimized by quantum particle swarm.
6. it is according to claim 1 based on the Electronic Nose feature extracting method for moving window function, it is characterised in that the window
Function is quarter window, Blackman window, hamming code window, Hanning window, rectangular window or Gaussian window.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510252261.7A CN104820074B (en) | 2015-05-18 | 2015-05-18 | Electronic nose characteristic extraction method based on moving window function |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510252261.7A CN104820074B (en) | 2015-05-18 | 2015-05-18 | Electronic nose characteristic extraction method based on moving window function |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104820074A CN104820074A (en) | 2015-08-05 |
| CN104820074B true CN104820074B (en) | 2017-05-10 |
Family
ID=53730429
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510252261.7A Expired - Fee Related CN104820074B (en) | 2015-05-18 | 2015-05-18 | Electronic nose characteristic extraction method based on moving window function |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104820074B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105675649B (en) * | 2016-01-11 | 2018-07-31 | 浙江大学 | A kind of electronic nose data characteristics extracting method based on adsorption dynamics adsorption kinetics |
| CN110940773A (en) * | 2019-11-22 | 2020-03-31 | 苏州慧闻纳米科技有限公司 | Processing method for response and recovery curve of odor in refrigerator |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103699446A (en) * | 2013-12-31 | 2014-04-02 | 南京信息工程大学 | Quantum-behaved particle swarm optimization (QPSO) algorithm based multi-objective dynamic workflow scheduling method |
| CN104572589A (en) * | 2015-01-12 | 2015-04-29 | 西南大学 | Electronic nose parameter synchronous optimization algorithm based on improved quantum particle swarm optimization algorithm |
-
2015
- 2015-05-18 CN CN201510252261.7A patent/CN104820074B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103699446A (en) * | 2013-12-31 | 2014-04-02 | 南京信息工程大学 | Quantum-behaved particle swarm optimization (QPSO) algorithm based multi-objective dynamic workflow scheduling method |
| CN104572589A (en) * | 2015-01-12 | 2015-04-29 | 西南大学 | Electronic nose parameter synchronous optimization algorithm based on improved quantum particle swarm optimization algorithm |
Non-Patent Citations (2)
| Title |
|---|
| 基于量子粒子群算法的聚类分析方法;叶安新 等;《计算机工程与应用》;20121111;第48卷(第32期);第52-55,97页 * |
| 面向伤口感染检测的电子鼻智能数据处理算法研究;贾鹏飞;《中国博士学位论文全文数据库工程科技Ⅱ辑》;20150415(第4期);C30-8 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104820074A (en) | 2015-08-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Xie et al. | A deep-learning-based real-time detector for grape leaf diseases using improved convolutional neural networks | |
| CN103488941B (en) | Hardware Trojan horse detection method and system | |
| CN109376574A (en) | Refuse to sentence radar HRRP target identification method based on CNN | |
| Rai et al. | Deep learning for high-impedance fault detection: Convolutional autoencoders | |
| CN103577755A (en) | Malicious script static detection method based on SVM (support vector machine) | |
| CN103134758B (en) | Rice leaf blast disease resistance identification grading method based on multi-scale hyperspectral image processing | |
| CN108846835A (en) | The image change detection method of convolutional network is separated based on depth | |
| CN106250848B (en) | A kind of single class Acoustic Object recognition methods based on multi-model fusion | |
| CN112735473B (en) | Method and system for identifying unmanned aerial vehicle based on voice | |
| CN110850244B (en) | Local discharge defect time domain map diagnosis method, system and medium based on deep learning | |
| CN104036289A (en) | Hyperspectral image classification method based on spatial and spectral features and sparse representation | |
| WO2023029390A1 (en) | Millimeter wave radar-based gesture detection and recognition method | |
| CN103776891A (en) | Method for detecting differentially-expressed protein | |
| CN104820074B (en) | Electronic nose characteristic extraction method based on moving window function | |
| CN109784383A (en) | A kind of rail cracks recognition methods based on figure characteristic of field and DS evidence theory fusion | |
| CN105300267B (en) | Plate thickness measuring method based on electromagnetic eddy and data anastomosing algorithm | |
| CN104125050B (en) | Ultrahigh frequency RFID card reader protocol conformance test method | |
| Aladhadh et al. | An efficient pest detection framework with a medium-scale benchmark to increase the agricultural productivity | |
| CN116434226A (en) | Circulating tumor cell analyzer | |
| CN105807204B (en) | Hardware Trojan horse detection method based on frequency spectrum refinement | |
| CN107169520A (en) | A kind of big data lacks attribute complementing method | |
| CN104749249B (en) | A kind of method of the detection physical purity of seed based on ultrasonic technology | |
| CN102914564A (en) | Detection method for rice kernel smut | |
| CN103268494A (en) | Parasite egg identifying method based on sparse representation | |
| CN109060933A (en) | The detection method of silver nano-grain content in soil |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| EXSB | Decision made by sipo to initiate substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170510 Termination date: 20210518 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |