Method for identifying matcha produced by different companies using fNIRS

Abstract

The application discloses a method for identifying matcha produced by different enterprises by using fNIRS, and belongs to the field of food detection. The method detects matcha produced by different enterprises by using functional near-infrared spectroscopy (fNIRS), carries out pretreatment and discriminant analysis on full-channel characteristic wave bands, screens characteristic channels, and further uses a discriminant model established by using characteristic wave bands of the characteristic channels to accurately identify matcha produced by different enterprises. Compared with traditional sensory evaluation methods and ingredient detection technologies, the method is objective and fair, does not need to train an evaluation personnel, time consumption is shortened, and the method can be used for rapidly identifying matcha produced by different enterprises.

Description

Technical Field

[0001] This invention relates to the field of food testing technology, and in particular to an fNIRS method for identifying matcha produced by different companies. Background Technology

[0002] Matcha, a type of green tea, has a long history and has long been loved by people all over the world, especially in China and Japan. It is reported that the consumption of matcha powder originated in 11th-century China, but it only became popular in Japan in the 14th century. In recent years, matcha has not only been consumed as a beverage but also added as a functional ingredient to other drinks and snacks, leading to continuous growth in the consumer market. However, along with the rapid development of the matcha industry, many problems have also emerged, such as unscrupulous merchants selling inferior products. This illegal activity seriously impacts the market, reduces consumer trust in matcha products, and greatly affects the healthy development of the matcha industry. Therefore, it is necessary to increase research on matcha authenticity assurance and brand protection technologies, using technological means to protect the rights of enterprises while ensuring consumers can consume matcha with peace of mind and safety.

[0003] Matcha produced by different companies varies in flavor and taste due to differences in raw materials, processing techniques, and storage methods. Traditional matcha flavor evaluation mainly relies on the physiological sensory evaluation of professionals after long-term training, but the results are inevitably affected by subjectivity. To make the measurement results more objective and stable, modern flavor analysis techniques (such as chromatography and mass spectrometry) have been introduced. However, these instruments also have some shortcomings: (1) The use of these instruments mainly relies on standard substance databases for qualitative and quantitative analysis, and due to the limitation of column polarity, they cannot completely detect flavor substances; (2) These instruments can only detect individual flavor compounds, but cannot detect the interactions between them (such as synergistic or inhibitory effects). Therefore, it is necessary to explore a fast and simple method for identifying matcha produced by different companies. Functional near-infrared spectroscopy (fNIRS) uses humans as the detection object and extracts taste signals from the cerebral cortex, which can quickly and objectively describe the overall flavor characteristics of food. fNIRS is not limited to characteristic compounds and does not require long-term training of evaluators, making it very suitable for companies to use in production. The fNIRS method was used to detect the flavor of matcha and to quickly identify matcha produced by different companies through specific data analysis methods. This has led to the initial establishment of a tea authenticity preservation technology for specific brands of matcha, which is of great significance in curbing the counterfeiting of matcha. Summary of the Invention

[0004] To overcome the shortcomings of existing technologies, the present invention aims to provide an fNIRS identification method for matcha produced by different companies. This method mainly extracts signals from the cerebral cortex of subjects using fNIRS technology, preprocesses and analyzes characteristic bands, filters characteristic channels, and establishes a discriminant model to accurately identify matcha produced by different companies. This method is objective, simple, fast, and accurate.

[0005] To achieve the above-mentioned objectives, the present invention provides the following technical solution:

[0006] This invention provides an fNIRS identification method for matcha produced by different companies, comprising the following steps:

[0007] Step 1: Collect and prepare matcha samples: Collect matcha from more than N different companies, take M parallel experimental samples from each company, take 0.5 to 2 grams of each matcha sample, place them in a tasting bowl and add ultrapure water at 70 to 90°C, stir at 500 to 1000 rpm for 20 to 40 seconds at room temperature to obtain matcha soup;

[0008] Step 2, Functional Near-Infrared Spectroscopy (fNIRS) detection: Subjects wear photoelectric caps, and the three-dimensional spatial position of each photoelectrode on the scalp of each subject is calibrated using a three-dimensional magnetic spatial digital instrument. The fNIRS channel is set to 2×12 channels. The matcha soup prepared in Step 1 is provided to the mouth of the matcha consumer through an automated liquid food oral sampling system. The time point when the matcha soup enters the oral cavity is taken as the starting point of the required wavelength information, and the swallowing time point is taken as the ending point of the required wavelength information. The fNIRS full-channel wavelength information of the matcha sample is recorded.

[0009] Step 3, fNIRS band information preprocessing: First, the intensity values ​​of the fNIRS full-channel band information obtained in Step 2 are converted into optical density signals using Beer-Lambert's law. Then, the motion artifacts are corrected by principal component analysis, and components with a cumulative variance greater than or equal to 80% are deleted. The corrected data is then filtered at 0.01 to 0.1 Hz and finally converted into oxyhemoglobin concentration values.

[0010] Step 4, Feature Channel Screening and Discriminant Model Establishment: Based on discriminant analysis, channels that have a significant impact on distinguishing matcha produced by different companies are screened out. The oxyhemoglobin concentration value of each channel is used as a feature to establish a discriminant model for accurately identifying matcha produced by different companies.

[0011] Preferably, the parameters of the automatic oral sampling system for liquid food described in step two are set as follows: 5-10 ml of matcha soup is injected into the mouth of the matcha consumer, the injection time is 3-8 seconds, the injection interval is 50-70 seconds, the matcha consumer holds the matcha soup in their mouth for 20-30 seconds and then swallows it, and the automatic oral sampling system for liquid food provides purified water to rinse the mouth.

[0012] Preferably, the motion artifact described in step three is a 3-5 second time period in which the standard deviation of the signal in a given channel changes by more than 10-15 within 0.5-1 second, and the amplitude exceeds 5-10.

[0013] In step four, the discriminant analysis can be any one of partial least squares discriminant analysis, linear discriminant analysis, or orthogonal partial least squares discriminant analysis; the discriminant model is a model established using the above discriminant analysis methods.

[0014] The present invention has the following beneficial effects:

[0015] (1) This invention uses a simple and fast fNIRS technology to analyze matcha produced by different companies. Without knowing the specific flavor substances, the overall characteristics of the matcha sample are determined by the taste signals of the subject's cerebral cortex. Furthermore, the screening channel modeling can accurately distinguish matcha from different brands.

[0016] (2) This invention does not require training of subjects, has a short detection time for each sample, is simple and easy to operate, does not require subjective evaluation by subjects, and the detection results are objective and reliable, and have high application value. Attached Figure Description

[0017] Figure 1 Flowchart of an fNIRS method for identifying matcha produced by different companies;

[0018] Figure 2 Partial least squares discriminant analysis score plots for matcha produced by different companies (CT, WT, and AB represent matcha samples from three companies);

[0019] Figure 3 The variable importance projection (VIP) score plot for 24 channels (red dots represent channels with VIP < 1);

[0020] Figure 4 Partial least squares discriminant score plot of matcha produced by different companies after channel screening;

[0021] Figure 5 A permutation test chart of the partial least squares discriminant model for matcha produced by different companies after channel screening. Detailed Implementation

[0022] The technical solutions provided by the present invention will be described in detail below with reference to the embodiments, but they should not be construed as limiting the scope of protection of the present invention.

[0023] Example 1

[0024] The experimental procedure is as follows Figure 1 As shown.

[0025] The matcha experimental samples totaled 9: CT (3), WT (3), and AB (3).

[0026] The initial recruitment of participants was based primarily on surveys regarding health, interests, communication skills, availability, attitudes toward matcha, and frequency of matcha consumption. A total of 7 people participated in the test (5 women and 2 men), aged between 20 and 35.

[0027] Experimental Method: 1g of each matcha sample was placed in a 500mL tasting bowl, and 75℃ ultrapure water was added. The mixture was stirred at 800r / min for 30s at room temperature (25±1℃) to obtain matcha infusion. 5mL of the matcha infusion was then administered to the subject via an automated oral liquid food delivery system. The injection time was 5s, the injection interval was 60s, and the injection pulse frequency was 6400Hz. The liquid food delivery tube and disposable liquid food delivery tube were pre-filled with the matcha solution. The disposable liquid food delivery tube was placed in the subject's mouth, and the matcha solution was placed in a liquid food container, while the purified water was placed in an oral cleaning solution container. Once the fNIRS imaging system begins recording, the participant clicks "Start" on the host computer's human-computer interaction system. The matcha solution is then automatically delivered into the subject's mouth via a peristaltic pump from a liquid food container. The process is paused after 5 seconds of injection, followed by a 30-second tasting, 30-second swallowing, and aftertaste evaluation. The system then automatically restarts, delivering a purified aqueous solution from a mouthwash container via a peristaltic pump. The subject repeats the swallowing and aftertaste evaluation. This completes one experiment. The experiment is then repeated three times. After each participant completes an experiment, the disposable liquid food delivery tube is replaced, and the entire experimental procedure is repeated until all participants have completed the experiment.

[0028] The fNIRS imaging system is a dual-band, portable 24-channel system, including an optical cap that completely covers the frontal lobe of the human body. Ten laser sources and eight optical probes are evenly distributed in a staggered pattern at 30mm intervals on the optical cap, with a data sampling rate of 50Hz. A three-dimensional magnetic spatial digitizer is used to measure the three-dimensional spatial position of each photoelectrode on the scalp of each subject. The fNIRS channel configuration is 2×12 channels, with Groups 1 and 2 set to DPF=6 and distance=30mm. The DPF setting is based on the formula 4.99+0.067×(Age^0.814), where Age ranges from 17 to 50.

[0029] Data Processing: First, abnormal channels and signals in the raw data were removed using data processing software. Then, the intensity values ​​were converted into optical density signals, and motion artifacts were corrected using PCA. Components with a cumulative variance greater than or equal to 80% were removed. Motion artifacts were defined as a 3-second time interval within which the standard deviation of the signal in a given active channel changed by more than 10 and the amplitude exceeded 5 within 0.5 seconds. The corrected data was then filtered from 0.01 to 0.1 Hz and converted into concentration values, with a partial path length factor of 6 for both wavelengths. The arithmetic mean of the filtered data was used for subsequent partial least squares discriminant analysis. Partial least squares discriminant analysis was performed on the 24 bands of the matcha data using the data processing software, identifying the 10... 9 The model is then enlarged and its absolute value is taken, followed by equal variance scaling. After discrimination, channels that are not important to the model, i.e., variables whose variable importance projection (VIP) is less than 1, are removed, and a new partial least squares discriminant model is built.

[0030] Results Analysis: In the partial least squares discriminant model established across all channels, the matcha produced by all three companies could be distinguished. Figure 2 ), and R 2 X, R 2 Y is greater than 0.5, while Q 2 The value is less than 0.5 (Table 1), so further model optimization is needed to stabilize the model. Channels with a VIP greater than 1 should be selected. Figure 3 Remodeling, its R 2 X, R 2 Y and Q 2 All three saw increases, and the matcha produced by the three companies can also be distinguished. Figure 4 Furthermore, after permutation testing, the model R under this permutation condition... 2 and Q 2 Highest( Figure 5 This demonstrates that the partial least squares discriminant model based on the VIP greater than 1 channel can accurately identify matcha produced by different companies.

[0031] Table 1. Partial Least Squares Discriminant Analysis Accuracy of Matcha Identification from Different Companies

[0032]

[0033] In this embodiment, all matcha samples on the market are examined, including samples of green tea powder masquerading as matcha powder. A discriminant model is established using fNIRS. When a matcha sample is introduced into the model, it can be determined whether it is genuine matcha and the brand can be identified. This method does not require senior matcha tasters or professional testing technicians; it is fast, convenient, and highly accurate, and is of great significance for the market promotion of matcha.

[0034] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An fNIRS method for identifying matcha produced by different companies, characterized in that, It includes the following steps: Step 1: Collect and prepare matcha samples: Collect matcha from more than N different companies, take M parallel experimental samples from each company, take 0.5~2 grams of each matcha sample, place them in a tasting bowl and add ultrapure water at 70~90℃, stir at 500~1000 rpm for 20~40 seconds at room temperature to obtain matcha soup; Step 2, Functional Near-Infrared Spectroscopy (fNIRS) detection: Subjects wear photoelectric caps, and the three-dimensional spatial position of each photoelectrode on the scalp of each subject is calibrated using a three-dimensional magnetic spatial digital instrument. The fNIRS channel is set to 2×12 channels. The matcha soup prepared in Step 1 is provided to the mouth of the matcha subject through an automated liquid food oral sampling system. The time point when the matcha soup enters the oral cavity is taken as the starting point of the required wavelength information, and the swallowing time point is taken as the ending point of the required wavelength information. The fNIRS full-channel wavelength information of the matcha sample is recorded. Step 3, fNIRS band information preprocessing: First, the intensity values ​​of the fNIRS full-channel band information obtained in Step 2 are converted into optical density signals using Beer-Lambert's law. Then, the motion artifacts are corrected by principal component analysis, and components with a cumulative variance greater than or equal to 80% are deleted. The corrected data is then filtered at 0.01~0.1Hz and finally converted into oxyhemoglobin concentration values. Step 4, Feature Channel Screening and Discriminant Model Establishment: Based on discriminant analysis, channels that have a significant impact on distinguishing matcha produced by different companies are screened out. The oxyhemoglobin concentration value of each channel is used as a feature to establish a discriminant model for accurately identifying matcha produced by different companies.

2. The fNIRS identification method for matcha produced by different companies according to claim 1, characterized in that, The parameters of the automatic oral sampling system for liquid food described in step two are set as follows: 5-10 ml of matcha soup is injected into the mouth of the matcha subject, the injection time is 3-8 seconds, the injection interval is 50-70 seconds, the matcha subject holds the matcha soup in his mouth for 20-30 seconds and then swallows it, and the automatic oral sampling system for liquid food provides purified water to rinse the mouth.

3. The fNIRS identification method for matcha produced by different enterprises according to claim 1, characterized in that, The motion artifacts mentioned in step three refer to the 3-5 second period when the standard deviation of the optical density signal of a given channel changes by more than 10 within 0.5-1 second and the amplitude exceeds 5.

4. The fNIRS identification method for matcha produced by different enterprises according to claim 1, characterized in that, The discriminant analysis described in step four is any one of partial least squares discriminant analysis, linear discriminant analysis, or orthogonal partial least squares discriminant analysis; the discriminant model is a model established using the discriminant analysis method.

Images