Beruflich Dokumente
Kultur Dokumente
Department of Biosystems Engineering, Zhejiang University, 886 Yuhangtang Road, Hangzhou 300058, China
College of Life Science and Engineering, Northwest University for Nationalities, Lanzhou 730024, China
a r t i c l e
i n f o
Article history:
Received 16 January 2013
Received in revised form 4 April 2013
Accepted 8 July 2013
Available online 15 July 2013
Keywords:
Electronic nose
Pattern recognition
Meat adulteration
Feature extraction
a b s t r a c t
The aims were to detect the adulteration of mutton by applying traditional methods (pH and color evaluation) and the E-nose, to build a model for prediction of the content of pork in minced mutton. An Enose of metal oxide sensors was used for the collection of volatiles presented in the samples. Feature
extraction methods, Principle component analysis (PCA), loading analysis and Stepwise linear discriminant analysis (step-LDA) were employed to optimize the data matrix. The results were evaluated by discriminant analysis methods, nding that step-LDA was the most effective method. Then Canonical
discriminant analysis (CDA) was used as pattern recognition techniques for the authentication of meat.
Partial least square analysis (PLS), Multiple Linear Regression (MLR) and Back propagation neural network (BPNN) were used to build a predictive model for the pork content in minced mutton. The model
built by BPNN could predict the adulteration more precisely than PLS and MLR do.
2013 Elsevier Ltd. All rights reserved.
1. Introduction
Adulteration of meat, involving the replacement of selected
breeds, particular geographical region or particular traditional
method with other cheaper animal proteins and even none meat
proteins (soy proteins), has attracted increasing attention. The
choice of one meat over another can reect the lifestyle, religion,
diet and health concerns. For example, lard, pork and meats not ritually slaughtered are forbidden for Muslims and Jews (Bonne and
Verbeke, 2008). However, cheaper animal protein, take pork as
an example, has been fraudulently used to substitute more expensive animal proteins, like mutton and beef. It requires reliable
methods for the authentication of meat adulteration.
Techniques that have been used in the detection of meat adulteration include molecular biology-based methods, enzyme linked
immunological methods, chromatographic methods and spectroscopy methods. Molecular biology-based methods, such as polymerase chain reaction (PCR), real-time PCR (Rodriguez et al.,
2005), restriction fragment length polymorphism analysis (RFLP)
(Chen et al., 2010), multiplex PCR (Ghovvati et al., 2009) and speciesspecic PCR (Man et al., 2007) have been used in the identication of species and adulteration of meat. Enzyme linked
immunological methods used in meat and meat products had been
reviewed by Asensio et al. (2008). These methods are the most specic and sensitive for species identication. However, they require
Corresponding author. Tel.: +86 571 88982178; fax: +86 571 88982191.
E-mail address: jwang@zju.edu.cn (J. Wang).
0260-8774/$ - see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jfoodeng.2013.07.004
745
illuminant and calibrated with a white plate. The means were used
for analysis.
2.3. The electronic nose (E-nose)
A PEN2 E-nose (portable electronic nose II, Airsense Corporation, Germany) was used to obtain the chemical ngerprint of
the samples. The basic system, which has been described in previous researches (Hai and Wang, 2006; Zhang et al., 2007), consisted
of a sampling apparatus, a detector unit containing the sensor array and pattern recognition software (Win Muster v.1.6) for data
recording and processing. The sensor array was composed of 10
different metal oxide sensors positioned into a small chamber.
Each sensor has a certain degree of afnity towards specic chemical or volatile compounds, and the nomenclature and characteristics of the sensors used are as follows: W1C (S1), sensitive to
aromatics; W5S (S2), sensitive to nitrogen oxides; W3C (S3), sensitive to ammonia, aromatic molecules; W6S (S4), sensitive to
hydrogen; W5C (S5), sensitive to methane, propane, and aliphatic
non-polar molecules; W1S (S6), sensitive to methane; W1W (S7),
sensitive to sulfur-containing organics; W2S (S8), sensitive to
broad alcohols; W2W (S9), sensitive to aromatics, sulfur- and chlorine-containing organics; W3S (S10), sensitive to methane and
aliphatic.
The experimental conditions for E-nose are given as follows:
10 g of the minced mixed meat was placed in a beaker of 250 ml
at the temperature of 25 C 3 C, and the beaker was sealed by
plastic for a headspace generation time of 30 min. The headspace
generation was carried out to increase the volatile compounds
from the meat sample. Before one sample was detected by E-nose,
the sensors were cleaned with the ow of fresh dry air, so that the
sample can be tested. Thereafter, the sensors were exposed to sample volatiles and the changes in sensors responses were acquired
by the data acquisition system (Winmuster). During the sampling
process, the sample gas was transferred into the sensor chamber
at a ow rate of 200 ml min1 and the collection time was 80 s at
an interval of 1 s.
746
Fig. 2. Color of the minced mutton adulterated with different content of pork.
747
Fig. 3. Ten sensors response curve to the minced mutton adulterated with different content of pork.
Table 1
The STEPDISC procedure stepwise selection.
Step
Number In
Entered
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
1
2
3
4
5
6
7
8
9
10
9
10
9
10
9
10
S5
S2-60
S6-60
S3
S1
S7-75
S3-60
S5-60
S8
S1-60
Removed
S3
S9-75
S1
S2-75
S2-60
S10-75
Partial R-square
F value
Pr > F
Wilks Lambda
Pr < Lambda
Pr > ASCC
0.9533
0.8707
0.4988
0.4076
0.6311
0.2225
0.213
0.2757
0.1605
0.1747
0.0603
0.1237
0.0635
0.119
0.0181
0.1133
465.35
152.23
22.29
15.28
37.64
6.24
5.84
8.15
4.05
4.45
1.35
2.97
1.42
2.84
0.39
2.68
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
0.0021
0.001
0.2505
0.0152
0.2218
0.0192
0.8563
0.0253
0.04670696
0.00603757
0.00302594
0.00179249
0.00066125
0.00051415
0.00040465
0.00029309
0.00024606
0.00020307
0.00021609
0.00018935
0.00020219
0.00017812
0.00018141
0.00016086
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
0.19065861
0.36440033
0.39701149
0.4549528
0.54188591
0.54814386
0.56269654
0.57913335
0.58464976
0.5921713
0.58448351
0.60806873
0.60708783
0.60942244
0.60843243
0.61918425
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
employed. The results obtained for each of the tasks were summarized in Table 2. It can be seen that the variable selection processes
afforded good results as regards to classication. The misclassied
were lower than 10.83% in most cases. Compared with data set
containing one point per sensor, 40 s, 60 s and 75 s, respectively,
the results obtained with data set containing three points per sensor were greatly improved especially for step-LDA, with one sample misclassied using CDA. This comparison showed that the
results were always improved through combining more
information.
The results obtained by different feature extraction methods
were also compared in Table 2. The best method was found to be
step-LDA for reducing the misclassied samples to the lowest
number of 1 using CDA, which might be explained by the elimination of the redundant signals of the original data set. However, the
results were worsened by loading analysis and PCA, as important
signals in discriminating the groups of meat were omitted. In
748
Table 2
Result of discriminant analysis.
40 s
60 s
75 s
Step-LDA
First 7 PCs
Loading analysis
Number of factors
10
10
10
30
10
11
14
10
8.33%
13
10.83%
13
10.83%
5
4.17%
4
3.33%
7
5.83%
8
6.67%
10
8.33%
13
10.83%
2
1.67%
1
0.83%
7
5.83%
10
8.33%
6
5.00%
7
5.83%
4
3.33%
5
4.17%
7
5.83%
7
5.83%
5
4.17%
Table 3
Comparison of three predictive models built on the responses of the E-nose optimized
by step-LDA.
Methods
PLS
MLR
BPNN
Calibration
Validation
R2
RMSEC (%)
R2
RMSEP (%)
0.9609
0.9609
0.9886
6.72
6.72
3.78
0.9092
0.91
0.9762
10.94
10.94
5.26
eters (Apetrei et al., 2007; Rudnitskaya et al., 2009). So, pork content in minced mutton could be determined simultaneously by
PLS using the E-nose data in this work.
The MLR algorithm establishes a model that describes the relationship between sensor signals and the pork content. The predictive models for pork content in minced mutton were given as
follow:
Large R and low RMSE indicate adequate ts. Table 3 showed the
prediction ability of the E-nose. Similar with the PLS, Table 3 illustrated a linear correlation between the responses of the sensors and
the pork content. When the model was used to predict the prediction set, the results were also high. The MLR model appeared to be
of high ability prediction for percentage of pork adulterated into
minced mutton with R2 higher than 0.93.
In conclusion, excellent prediction ability of E-nose was found
using PLS, MLR and BPNN for prediction of pork content in minced
mutton with R2 higher than 0.9092 and RMSE lower than 10.94%.
The best results were found by BPNN model, with high correlation
(higher than 0.97) for the training and testing subsets.
4. Conclusions
In this paper, an E-nose was used to detect the adulteration of
pork in minced mutton. Multivariate analysis methods were employed to explore the performance of the E-nose in classication
of the adulteration. 120 samples were detected and the signals
were analyzed by three feature extraction methods and pattern
recognition techniques. Step-LDA was proved to be the most effective feature extraction method. It is applicable for the E-nose to detect the adulteration in mutton based on the Step-LDA using CDA.
MLR, PLS and BPNN were used to predict the pork content in
minced mutton precisely. Three methods showed high capacity
in prediction for content of pork in minced mutton. Whats more,
BPNN was the most effective method for the prediction of pork
content. The E-nose had been proved to be a useful authentication
method for meat adulteration detection for its efciency and high
accuracy.
Acknowledgements
The authors acknowledge the nancial support of the Chinese
National Foundation of Nature and Science through Project
31071548,
National
Key
Technology
R&D
Program
2012BAD29B02-4, the Research Fund for the Doctoral Program
of Chinese National Higher Education through Project
20100101110133, and the Fundamental Research Funds for the
Central Universities through Project ZYZ2012071.
References
Apetrei, C., Apetrei, I.M., Nevares, I., del Alamo, M., Parra, V., Rodrguez-Mndez,
M.L., De Saja, J.A., 2007. Using an e-tongue based on voltammetric electrodes to
discriminate among red wines aged in oak barrels or aged using alternative
methods: correlation between electrochemical signals and analytical
parameters. Electrochimica Acta 52 (7), 25882594.
749
Asensio, L., Gonzlez, I., Garca, T., Martin, R., 2008. Determination of food
authenticity by enzyme-linked immunosorbent assay (ELISA). Food Control 19
(1), 18.
Bonne, K., Verbeke, W., 2008. Muslim consumer trust in halal meat status and
control in Belgium. Meat Science 79 (1), 113123.
Chen, S.-Y., Liu, Y.-P., Yao, Y.-G., 2010. Species authentication of commercial beef
jerky based on PCRRFLP analysis of the mitochondrial 12S rRNA gene. Journal
of Genetics and Genomics 37 (11), 763769.
Chou, C.C., Lin, S.P., Lee, K.M., Hsu, C.T., Vickroy, T.W., Zen, J.M., 2007. Fast
differentiation of meats from fteen animal species by liquid chromatography
with electrochemical detection using copper nanoparticle plated electrodes.
Journal of Chromatography B 846 (12), 230239.
Cosio, M.S., Ballabio, D., Benedetti, S., Gigliotti, C., 2006. Geographical origin and
authentication of extra virgin olive oils by an electronic nose in combination
with articial neural networks. Analytica Chimica Acta 567 (2), 202210.
Cozzolino, D., Murray, I., 2004. Identication of animal meat muscles by visible and
near infrared reectance spectroscopy. LWT Food Science and Technology 37
(4), 447452.
Gmez, A.H., Wang, J., Hu, G., Pereira, A.G., 2006. Electronic nose technique potential
monitoring mandarin maturity. Sensors and Actuators B: Chemical 113 (1),
347353.
Gmez, A.H., Wang, J., Hu, G., Pereira, A.G., 2007. Discrimination of storage shelf-life
for mandarin by electronic nose technique. LWT Food Science and Technology
40 (4), 681689.
Garca, M., Aleixandre, M., Gutirrez, J., Horrillo, M.C., 2006. Electronic nose for ham
discrimination. Sensors and Actuators B: Chemical 114 (1), 418422.
GB/T9695.5, 2008. National Standard of the Peoples Republic of China, Meat and
Meat ProductsMeasurement of pH.
Ghovvati, S., Nassiri, M.R., Mirhoseini, S., Moussavi, A.H., Javadmanesh, A., 2009.
Fraud identication in industrial meat products by multiplex PCR assay. Food
Control 20 (8), 696699.
Haddi, Z., Amari, A., Ali, A.O., Bari, N.E., Barhoumi, H., Maaref, A., Jaffrezic-Renault,
N., Bouchikhi, B., 2011. Discrimination and identication of geographical origin
virgin olive oil by an e-nose based on MOS sensors and pattern recognition
techniques. Procedia Engineering 25, 11371140.
Hai, Z., Wang, J., 2006. Electronic nose and data analysis for detection of maize oil
adulteration in sesame oil. Sensors and Actuators B: Chemical 119 (2), 449455.
Limbo, S., Torri, L., Sinelli, N., Franzetti, L., Casiraghi, E., 2010. Evaluation and
predictive modeling of shelf life of minced beef stored in high-oxygen modied
atmosphere packaging at different temperatures. Meat Science 84 (1), 129136.
Man, Y.B.C., Aida, A.A., Raha, A.R., Son, R., 2007. Identication of pork derivatives in
food products by speciesspecic polymerase chain reaction (PCR) for halal
verication. Food Control 18 (7), 885889.
Man, Y.B.C., Gan, H.L., NorAini, I., Nazimah, S.A.H., Tan, C.P., 2005. Detection of lard
adulteration in RBD palm olein using an electronic nose. Food Chemistry 90 (4),
829835.
Meza-Marquez, O.G., Gallardo-Velazquez, T., Osorio-Revilla, G., 2010. Application of
mid-infrared spectroscopy with multivariate analysis and soft independent
modeling of class analogies (SIMCA) for the detection of adulterants in minced
beef. Meat Science 86 (2), 511519.
Musatov, V.Y., Sysoev, V.V., Sommer, M., Kiselev, I., 2010. Assessment of meat
freshness with metal oxide sensor microarray electronic nose: a practical
approach. Sensors and Actuators BChemical 144 (1), 99103.
Nan, Q., 2003. Manual for Meat Industry, rst ed. China Light Industry Press,
Beijing, China, pp. 7274.
Nurjuliana, M., Che Man, Y.B., Mat Hashim, D., Mohamed, A.K.S., 2011. Rapid
identication of pork for halal authentication using the electronic nose and gas
chromatography mass spectrometer with headspace analyzer. Meat Science 88
(4), 638644.
Penza, M., Cassano, G., 2004. Recognition of adulteration of Italian wines by thinlm multisensor array and articial neural networks. Analytica Chimica Acta
509 (2), 159177.
Rodriguez, M.A., Garcia, T., Gonzalez, I., Hernandez, P.E., Martin, R., 2005. TaqMan
real-time PCR for the detection and quantitation of pork in meat mixtures. Meat
Science 70 (1), 113120.
Rohman, A., Sismindari, Erwanto, Y., Man, Y.B.C., 2011. Analysis of pork adulteration
in beef meatball using Fourier transform infrared (FTIR) spectroscopy. Meat
Science 88 (1), 9195.
Rudnitskaya, A., Polshin, E., Kirsanov, D., Lammertyn, J., Nicolai, B., Saison, D.,
Delvaux, F.R., Delvaux, F., Legin, A., 2009. Instrumental measurement of beer
taste attributes using an electronic tongue. Analytica Chimica Acta 646 (12),
111118.
Sacco, D., Brescia, M.A., Buccolieri, A., Jambrenghi, A.C., 2005. Geographical origin
and breed discrimination of Apulian lamb meat samples by means of analytical
and spectroscopic determinations. Meat Science 71 (3), 542548.
Yu, H., Wang, J., 2007. Discrimination of LongJing green-tea grade by electronic
nose. Sensors and Actuators B: Chemical 122 (1), 134140.
Zhang, H., Wang, J., Tian, X., Yu, H., Yu, Y., 2007. Optimization of sensor array and
detection of stored duration of wheat by electronic nose. Journal of Food
Engineering 82 (4), 403408.
Zhang, N., Zhang, D., Li, S., Li, Q., 2008. Preliminary study on origin traceability of
mutton by near infrared reectance spectroscopy coupled with SIMCA method.
Transactions of the Chinese Society of Agricultural Engineering 24 (12), 309
312.