Limits...
Recognition of FT-IR Data Cuscutae Semen, Japanese Dodder, and Sinapis Semen Using Discrete Wavelet Transformation and RBF Networks.

Hu T, Weng X, Xu L, Cheng C, Yu P - J Anal Methods Chem (2013)

Bottom Line: Thus five feature parameters form the feature vector.The feature vector is input to the RBF neural networks to train so as to accurately classify the cuscutae semen, Japanese dodder, and sinapis semen. 120 sets of FT-IR data are used to train and test the proposed method, where 60 sets of data are used to train samples, and another 60 sets of FT-IR data are used to test samples.Experimental results show that the accurate recognition rate of cuscutae semen, Japanese dodder, and sinapis semen is average of 100.00%, 98.33%, and 100.00%, respectively, following the proposed method.

View Article: PubMed Central - PubMed

Affiliation: Faculty of Life Science and Chemical Engineering, Huaiyin Institute of Technology, Huaian 223003, China ; National Special Superfine Powder Engineering Center, Nanjing University of Science and Technology, Nanjing 210094, China.

ABSTRACT
Horizontal attenuation total reflection Fourier transformation infrared spectroscopy (HATR-FT-IR) studies on cuscutae semen and its confusable varieties Japanese dodder and sinapis semen combined with discrete wavelet transformation (DWT) and radial basis function (RBF) neural networks have been conducted in order to classify them. DWT is used to decompose the FT-IRs of cuscutae semen, Japanese dodder, and sinapis semen. Two main scales are selected as the feature extracting space in the DWT domain. According to the distribution of cuscutae semen, Japanese dodder, and sinapis semen's FT-IRs, three feature regions are determined at detail 3, and two feature regions are determined at detail 4 by selecting two scales in the DWT domain. Thus five feature parameters form the feature vector. The feature vector is input to the RBF neural networks to train so as to accurately classify the cuscutae semen, Japanese dodder, and sinapis semen. 120 sets of FT-IR data are used to train and test the proposed method, where 60 sets of data are used to train samples, and another 60 sets of FT-IR data are used to test samples. Experimental results show that the accurate recognition rate of cuscutae semen, Japanese dodder, and sinapis semen is average of 100.00%, 98.33%, and 100.00%, respectively, following the proposed method.

No MeSH data available.


Related in: MedlinePlus

Structure of RBF neural network.
© Copyright Policy - open-access
Related In: Results  -  Collection


getmorefigures.php?uid=PMC3824338&req=5

fig1: Structure of RBF neural network.

Mentions: RBF neural network can extend or preprocess the input vector to the high-dimensional space. It not only has good generalization ability and also avoids the complex computation as back-propagation neural network. Therefore we can achieve the rapid learning of neural network. In this paper, we aim at the classification and identification of three kinds of plant seeds (cuscutae semen, Japanese dodder and sinapis semen). Five feature parameters are used as input vector, thus the input layer of the network needs five neurons. Therefore the RBF neural network has five input neural units and three output neural units. Structure of RBF neural network is shown in Figure 1.


Recognition of FT-IR Data Cuscutae Semen, Japanese Dodder, and Sinapis Semen Using Discrete Wavelet Transformation and RBF Networks.

Hu T, Weng X, Xu L, Cheng C, Yu P - J Anal Methods Chem (2013)

Structure of RBF neural network.
© Copyright Policy - open-access
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC3824338&req=5

fig1: Structure of RBF neural network.
Mentions: RBF neural network can extend or preprocess the input vector to the high-dimensional space. It not only has good generalization ability and also avoids the complex computation as back-propagation neural network. Therefore we can achieve the rapid learning of neural network. In this paper, we aim at the classification and identification of three kinds of plant seeds (cuscutae semen, Japanese dodder and sinapis semen). Five feature parameters are used as input vector, thus the input layer of the network needs five neurons. Therefore the RBF neural network has five input neural units and three output neural units. Structure of RBF neural network is shown in Figure 1.

Bottom Line: Thus five feature parameters form the feature vector.The feature vector is input to the RBF neural networks to train so as to accurately classify the cuscutae semen, Japanese dodder, and sinapis semen. 120 sets of FT-IR data are used to train and test the proposed method, where 60 sets of data are used to train samples, and another 60 sets of FT-IR data are used to test samples.Experimental results show that the accurate recognition rate of cuscutae semen, Japanese dodder, and sinapis semen is average of 100.00%, 98.33%, and 100.00%, respectively, following the proposed method.

View Article: PubMed Central - PubMed

Affiliation: Faculty of Life Science and Chemical Engineering, Huaiyin Institute of Technology, Huaian 223003, China ; National Special Superfine Powder Engineering Center, Nanjing University of Science and Technology, Nanjing 210094, China.

ABSTRACT
Horizontal attenuation total reflection Fourier transformation infrared spectroscopy (HATR-FT-IR) studies on cuscutae semen and its confusable varieties Japanese dodder and sinapis semen combined with discrete wavelet transformation (DWT) and radial basis function (RBF) neural networks have been conducted in order to classify them. DWT is used to decompose the FT-IRs of cuscutae semen, Japanese dodder, and sinapis semen. Two main scales are selected as the feature extracting space in the DWT domain. According to the distribution of cuscutae semen, Japanese dodder, and sinapis semen's FT-IRs, three feature regions are determined at detail 3, and two feature regions are determined at detail 4 by selecting two scales in the DWT domain. Thus five feature parameters form the feature vector. The feature vector is input to the RBF neural networks to train so as to accurately classify the cuscutae semen, Japanese dodder, and sinapis semen. 120 sets of FT-IR data are used to train and test the proposed method, where 60 sets of data are used to train samples, and another 60 sets of FT-IR data are used to test samples. Experimental results show that the accurate recognition rate of cuscutae semen, Japanese dodder, and sinapis semen is average of 100.00%, 98.33%, and 100.00%, respectively, following the proposed method.

No MeSH data available.


Related in: MedlinePlus