Summary This Resistance Band Set can be integrated seamlessly with every popular workout program including Yoga, Pilates, and more. Or use them for general exercise, stretching, strength training, power weight programs.These Resistance Bands are often used for sports and fitness, physical therapists love these physical therapy bands to help them rehabilitate their patients. Our Stretch Bands work for people suffering from leg, knee and back injuries. Fitness Resistance Bands,Custom Resistance Exercise Band,Power Jump Resistance Band,Nylon Custom Resistance Exercise Bands Shandong Shangkun Sports Health Technology Co.,Ltd. , http://www.shangkunsports.com
Rhubarb is one of the most commonly used Chinese herbal medicines in China. The rapid and accurate identification of authentic and non-authentic rhubarb is of great significance for the quality control of rhubarb and its herbal products. Combining near infrared diffuse reflectance spectroscopy analysis technology with artificial neural network method, 52 kinds of rhubarb samples were measured and identified, and the accuracy rate was up to 96%. The effects of the number of hidden layers and momentum factors of the neural network are discussed. NIR spectroscopy has the characteristics of less sample preparation, rapid determination and non-destructive determination, so it is particularly suitable for the identification of Chinese herbal medicine.
Key words: near infrared spectroscopy; rhubarb; artificial neural network; Chinese herbal medicine
introduction
Rhubarb is a plant belonging to the genus Rheum of the Polygonaceae family, with a total of about 60 species. Rhubarb is a famous specialty medicinal material in China and has a long history of medication. Rhubarb has the effects of diarrhea, antibacterial, gall bladder, liver protection, hemostasis and blood circulation. The roots and rhizomes of non-genuine rhubarb are often mixed with commodities, but their diarrhea is not as effective as genuine rhubarb, and some can cause abdominal pain. In order to ensure the clinical efficacy of genuine rhubarb, for a long time, the identification of rhubarb crude drugs mostly depends on their external morphology, character identification, microscopic identification and physical and chemical identification. These methods rely on experience to a certain extent, and it is difficult to distinguish genuine rhubarb from non-genuine The root of authentic rhubarb and the processed powder. In recent years, people have used a variety of modern analysis and detection techniques to study the identification of various Chinese herbal medicines, such as infrared spectroscopy and mass spectrometry. Using these modern analysis and detection methods makes it possible to quickly and accurately identify Chinese herbal medicines, and also lays a scientific foundation for the modernization of Chinese medicine.
Near-infrared spectroscopy (NIR) technology is one of the spectral analysis methods that has received special attention and rapid development in recent years. In the near infrared spectral region (usually 780 ~ 3100nm), it is mainly the frequency doubling and combined frequency absorption of CH, OH and NH groups in the molecule. Due to its weak absorption band, a large sample size is often required to obtain an ideal spectrum. Because the overlap of spectral peaks in the NIR region is very serious, data processing and interpretation are very difficult. In the earlier period, due to the limitation of technical level and experimental conditions, the information in the NIR area could not be fully presented.
Therefore, the application of near infrared spectroscopy is very limited. With the popularization of computer technology and the development of chemometrics, people have made in-depth research on the interpretation and calibration of NIR spectral data. Many calibration methods for NIR spectral analysis have been established, which makes NIR spectral analysis technology widely used. Current near infrared spectroscopy
The method has a wide range of applications in the fields of agriculture, food industry, petroleum, chemical, pharmaceutical, textile and biomedicine, especially for online analysis. There are also many reports on the application of near infrared spectroscopy in the identification of Chinese herbal medicine. However, there are few reports on the application of Chinese herbal medicine rhubarb.
Artificial neural network (ANN) is a mathematical simulation of biological neural network. It uses a large number of simple processing units to connect a complex network composed of a wide range, to imitate the structure and function of the human brain neural network, so as to process information. Since neural networks have functions such as self-organization, self-learning, robustness, fault tolerance, and nonlinear information processing, they have been widely used in various fields. At present, the most widely used is the feedforward neural network (BP-ANN) using error back propagation algorithm.
In this paper, the combination of near infrared spectroscopy and artificial neural network method, using diffuse reflection optical detection method, 52 kinds of rhubarb samples were measured and identified, and the influence of the number of hidden layers and momentum factors of the neural network were discussed.
1 Experimental part
1.1 Instruments and samples
Foss 6500 near infrared spectrometer (Foss NIR Systems Inc., MD, USA), quartz halogen lamp, PbS detector. The 52 rhubarb samples selected in this work are samples of different varieties and different origins. According to the requirements of China's Pharmacopoeia, these samples are divided into two categories: authentic rhubarb and non-authentic rhubarb, of which 25 are authentic samples (No. 1-25) and 27 are non-authentic samples (No. 25-52). The rhubarb sample is dried and crushed into a 60-mesh powder for direct measurement.
1.2 Data collection and processing
The measurement wavelength range of rhubarb samples is: 1100 ～ 2500nm. Collect a data point every 2nm. The sample cell used for spectrum acquisition is 38 mm in diameter and 10 mm in thickness. In order to ensure the representativeness of the sample data, the sample cell is taken out and shaken after several measurements to refill the sample in the sample cell. Each sample was scanned 50 times, and then the average value was taken as the spectrum of the sample. The scanned measurement data of the sample is stored in ASCII code, and then calculated and processed by another microcomputer.
The measured NIR spectrum of the rhubarb sample is processed by the second derivative to eliminate the influence of the slope background. In order to reduce the spectral variables and thus improve the training speed of the neural network, we use the wavelet transform method to compress the second derivative N1R spectrum. After wavelet compression, the spectral variable points are reduced from 700 to 44. The use of wavelet transform data compression technology can not only efficiently reduce the number of data variables, but also maintain the characteristics of the original spectrum. The compressed rhubarb NIR spectrum with 44 variables was used as the input of the neural network. In this work, the neural network of error back propagation algorithm (BP-ANN) is used to establish the classification model of rhubarb samples. The input layer unit of the neural network is 44, and the output layer unit is 1, with 1.0 representing genuine rhubarb and 0.0 representing non-genuine rhubarb. Optimize the selection of hidden layer units.
In order to simplify the calculation steps, first use Matlab 5.0 (Mathworks, Inc., USA) internal function Appooef to perform one-dimensional wavelet transform to compress the spectrum. Then use Trmnbpx (fast BP algorithm) for network training and modeling. In order to verify the classification model established by the neural network, a cross-validation method is adopted. Use the method of 1 in n to select the test sample, that is, select one sample at a time as the test sample, and the remaining samples as the training samples. In this way, each sample is used as a test sample once and n times as a training set sample. The judgment threshold of the prediction result is set to 0.5, that is, when the output value is greater than 0.5, it is judged as genuine rhubarb, and when the output value is less than 0.5, it is judged as non-genuine rhubarb.
2 Results and discussion
2.1 NIR spectrum and similarity discrimination of rhubarb
It can be seen that the near-infrared spectra of rhubarb samples are very similar, and it is impossible to distinguish genuine and non-genuine rhubarb by direct observation. Even using the traditional correlation coefficient method, it is difficult to identify authentic and non-authentic rhubarb samples. The correlation coefficients of authentic and non-authentic rhubarb samples are very large (ie very close to 1). Even if the derivative spectrum is used to reduce the correlation coefficient between genuine and non-genuine samples, the rhubarb samples cannot be classified and identified based on NIR spectroscopy. Therefore, we have conducted in-depth research on the classification and identification of genuine and non-genuine rhubarb samples using artificial neural network methods.
2.2 The influence of hidden layer nodes
In this experiment, we used BP-ANN to distinguish genuine rhubarb from non-genuine rhubarb. In fact, the number of nodes in the hidden layer determines the complexity of the BP network. Therefore, an optimal number of hidden layer nodes must be selected. We compare the number of hidden layers from 1 to 8 one by one. It can be found that when the hidden layer neuron is 1, the BP network's recognition rate of rhubarb is very low, only reaching 55.8%. When the hidden layer neuron is 2, the BP network's recognition rate of rhubarb immediately increases to 90.38%. When the hidden layer neurons are adjusted to 2 to 8, the BP network's recognition accuracy of rhubarb tends to be flat. When the hidden layer neurons is 5, the BP network's recognition accuracy of rhubarb reaches the highest, which is 96.15%. After the above optimization, we selected the best number of hidden layers to be 5.
2.3 The influence of momentum factor
Momentum factor and learning rate are two important factors that affect the training rate and convergence of BP neural network. So far, there are no strict systematic theoretical rules to choose the momentum factor and learning rate. For specific problems, these parameters are usually selected based on experiments. In this experiment, we used the internal function Trainbpx in Matlab software, where the learning rate was adjusted by the function itself. Therefore, only the appropriate momentum factor needs to be selected in the experiment. We compare the momentum factors from 0.1 to 0.9 one by one, and finally, we determine the most suitable momentum factor to be 0.9. The BP network's recognition accuracy rate of rhubarb is relatively stable with the change of momentum factor, which has reached more than 90%. Adjusting the momentum factor allows us to find the highest recognition accuracy, but if the momentum factor is too large, the BP network cannot converge.
It can be found that there are two rhubarb samples that are wrongly identified outside the BP network. The output value of the second sample is 0.4938, and the output value of the 36th sample is 0.7228. These two samples are the samples that have been erroneously identified. The identification of the other 50 samples is correct. On the whole, the identification accuracy of using BP network to identify rhubarb samples reached 96.15%.
3 Conclusion
The purpose of this article is to combine near infrared spectroscopy and artificial neural network for the identification of traditional Chinese medicine rhubarb. After the rhubarb sample is crushed, it can be analyzed and monitored by near infrared spectrometer without complicated processing. After the NIR spectrum is compressed by wavelet transform, the spectrum variable is reduced from 700 to 44. A wavelet-compressed NIR spectrum is input to the neural network to establish a classification and identification model. Using independent prediction samples for testing, the recognition accuracy rate can reach 96.15%. The method is easy to operate, has no pollution and low consumption, and is a promising method for identifying Chinese herbal medicine.
Study on near infrared spectroscopy of Chinese herbal medicine rhubarb