Sensory evaluation method and system

Abstract

The invention discloses a sensory evaluation method and system, and the method comprises the steps: outputting a form input interface corresponding to a specified evaluation method in response to a specified instruction of the evaluation method received from a user interaction interface; responding to form input data received from the form input interface, and automatically running a specified evaluation method; and outputting a running result of the specified evaluation method through the user interaction interface. According to the method, automatic sensory evaluation is quickly started through a convenient operation mode, a user does not need to write a statistical script by himself or herself, manual calculation and result analysis are not needed, the analysis efficiency is improved, and the reliability of an analysis result and the user experience are remarkably improved.

Description

Technical Field

[0001] This application relates to the technical field of light-sensitive evaluation, and more specifically, to a sensory evaluation method and system. Background Art

[0002] The A-not-A sensory evaluation method is a statistical method commonly used in the field of sensory analysis, mainly used to determine whether there are significant sensory differences between two types of samples. Currently, the mainstream implementation of A-not-A sensory evaluation relies on manual calculation or the use of general programming languages (such as R, Python, etc.).

[0003] For example, the Chinese standard GB / T 39558-2020 "Methodology of Sensory Analysis - 'A'-'not A' Test" stipulates the methodology of the A-not-A sensory evaluation method, clarifying the steps of the A-not-A test, the selection of assessors, and the data analysis method. However, this standard only provides guidance for the methodology and does not involve the development of specific tools or systems. Therefore, although the existing standard provides a method framework for sensory analysis, manual calculation and traditional statistical tools are still required in actual operation.

[0004] Therefore, the traditional implementation has the following problems in practical applications:

[0005] 1. Low efficiency: All steps such as data input, calculation, and result parsing need to be manually operated, resulting in low analysis efficiency.

[0006] 2. Complex operation: The existing methods require users to write statistical scripts by themselves, which has a high technical threshold for non-professional users.

[0007] 3. Poor interactivity: Existing analysis tools usually lack an intuitive user interface, and users need to operate through command lines or complex configuration files, resulting in a poor operation experience. Summary of the Invention

[0008] This application provides a sensory evaluation method and system, which can quickly start automatic sensory evaluation through a convenient operation method. Users do not need to write statistical scripts by themselves, nor do they need to perform manual calculation and result parsing, improving the analysis efficiency and significantly enhancing the reliability of the analysis results and the user experience.

[0009] This application provides a sensory evaluation method, including:

[0010] Responding to a specified instruction of the evaluation method received from the user interface, outputting a form input interface corresponding to the specified evaluation method;

[0011] Responding to the form input data received from the form input interface, automatically running the specified evaluation method;

[0012] Output the operation result of the specified evaluation method through the user interaction interface.

[0013] Preferably, after receiving the form input data from the form input interface, verify the form input data; if the verification passes, automatically run the specified evaluation method.

[0014] Preferably, if the specified evaluation method is the A-non-A sensory evaluation method, when automatically running the specified evaluation method, it includes:

[0015] Calculate the sensory difference intensity;

[0016] Calculate the significance of the difference between samples.

[0017] Preferably, if the specified evaluation method is the A-non-A sensory evaluation method, the verification of the form input data specifically includes:

[0018] Judge whether all form input data are positive integers;

[0019] If so, judge whether the number of samples determined as "A" in the A samples exceeds the total number of A samples, and judge whether the number of samples determined as "A" in the non-A samples exceeds the total number of A samples;

[0020] If neither exceeds, the verification passes.

[0021] Preferably, calculating the sensory difference intensity includes:

[0022] Construct a Probit regression model of the binomial distribution;

[0023] Obtain the regression coefficients of the Probit regression model based on maximum likelihood estimation, and the regression coefficients include the first intercept and the first slope;

[0024] Take the opposite of the first slope as the sensory difference intensity.

[0025] Preferably, obtaining the regression coefficients of the Probit regression model based on maximum likelihood estimation includes:

[0026] Obtain the design matrix corresponding to the form input data based on the Probit regression model;

[0027] Construct a linear regression function based on the design matrix, where the intercept and slope of the linear regression function are the first intercept and the first slope;

[0028] Iterate on the first intercept and the first slope based on the maximum likelihood function, and after the iteration is completed, obtain the predicted value of the first intercept and the predicted value of the first slope.

[0029] Preferably, iterating on the first intercept and the first slope based on the maximum likelihood function includes:

[0030] In each iteration, calculate the first score of the first intercept and the second score of the first slope based on the maximum likelihood function, and construct the Hessian matrix based on the maximum likelihood function; then update the first intercept and the first slope based on the initial first intercept and the initial first slope, the first score, the second score, and the Hessian matrix in this iteration, and use the updated first intercept and the first slope as the initial first intercept and the initial first slope in the next iteration;

[0031] If the convergence condition is satisfied, use the first intercept and the first slope in the last iteration as the predicted value of the first intercept and the predicted value of the first slope respectively.

[0032] Preferably, constructing the Hessian matrix includes:

[0033] Calculate the first contribution matrix for sample type "A" and the second contribution matrix for sample type "non - A". The first contribution matrix and the second contribution matrix are two - row and two - column matrices. Determine the second intercept and the third intercept according to the first column of the first contribution matrix and the second contribution matrix respectively, and determine the second slope and the third slope according to the second column of the first contribution matrix and the second contribution matrix respectively;

[0034] Calculate the four element values of the two - row and two - column Hessian matrix using the second intercept, the third intercept, the second slope, and the third slope.

[0035] This application also provides a sensory evaluation system, including a user interaction interface and a background processing module. The background processing module is used to execute the sensory evaluation method according to claims 1 - 7.

[0036] Preferably, the background processing module is encapsulated as a dynamic link library.

[0037] Through the following detailed description of the exemplary embodiments of the present application with reference to the accompanying drawings, other features and advantages of the present application will become clear. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments of the present application and, together with the description, serve to explain the principles of the present application.

[0039] Figure 1 It is a structural diagram of the sensory evaluation system provided by the present application;

[0040] Figure 2 It is a flowchart of the sensory evaluation method provided by the present application;

[0041] Figure 3 It is a schematic diagram of an embodiment of the form input interface provided by the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0042] Various exemplary embodiments of the present application will now be described in detail with reference to the accompanying drawings. It should be noted that: unless otherwise specifically stated, the relative arrangements of components and steps, numerical expressions, and numerical values set forth in these embodiments do not limit the scope of the present application.

[0043] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way a limitation on the present application or its application or use.

[0044] Technologies, methods, and devices known to those of ordinary skill in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and devices should be regarded as part of the specification.

[0045] In all the examples shown and discussed herein, any specific values should be construed as merely exemplary and not as a limitation. Thus, other examples of the exemplary embodiments may have different values.

[0046] The present application provides a sensory evaluation method and system. The integrated sensory evaluation system based on the PC side can quickly start automatic sensory evaluation through a convenient operation method, simplifies the operation process, and users do not need to write statistical scripts by themselves, nor do they need to perform calculations and result analysis manually, improving the analysis efficiency and significantly enhancing the reliability of the analysis results and the user experience.

[0047] As Figure 1 shown, the sensory evaluation system provided by the present application includes a user interaction interface and a background processing module. The user interaction interface is used to interact with the user, and the user interaction interface is used to receive the user's instructions and output the operation results. The background processing module is used to automatically run the specified evaluation method according to the data input by the user. The background processing module integrates the processing programs of various types of sensory evaluation methods, such as integrating various types of sensory evaluation methods such as difference evaluation, similarity evaluation, and comprehensive evaluation into the same system to meet the needs of different industries.

[0048] As an embodiment, the user interaction interface is developed using the Tcl / Tk or Qt framework, supporting cross-platform operation to ensure stable operation in operating systems such as Windows and MacOS.

[0049] Preferably, the background processing module is encapsulated as a dynamic link library (DLL) to improve the calculation efficiency and system response speed, achieving a millisecond-level calculation response. The complete process from data input to result output takes less than 5 seconds, and the efficiency is improved by more than 90% compared with traditional manual calculation or script operation.

[0050] Based on the above sensory evaluation system, the present application also provides a sensory evaluation method executed by the background processing module. As Figure 2 shown, the sensory evaluation method includes:

[0051] S210: In response to receiving a specified instruction for an evaluation method from a user interface, output a form input interface corresponding to the specified evaluation method.

[0052] On the user interface, there are entrances for the user to select the type of sensory evaluation and the sensory evaluation method, such as selecting the "A-not-A evaluation method" in the differential evaluation.

[0053] S220: In response to receiving form input data from the form input interface, automatically run the specified evaluation method.

[0054] For the form input interface corresponding to the "A-not-A evaluation method", the determination results of each group of samples can be input, and each group includes the sample type (A sample or non-A sample), the number of samples determined as "A" in this group of samples (marked as z i ) and the total number of samples in this group of samples (marked as n i ).

[0055] As an example, as Figure 3 shown, the data input by the user includes two groups:

[0056] The first group is A samples, the number of samples determined as "A" is 10, and the total number of samples in this group is 20;

[0057] The second group is non-A samples, the number of samples determined as "A" is 3, and the total number of samples in this group is 20.

[0058] S230: Output the operation result of the specified evaluation method through the user interface.

[0059] Preferably, after receiving the form input data from the form input interface, the post-processing module also verifies the form input data to avoid incorrect data affecting the analysis result; if the verification passes, the specified evaluation method is automatically run.

[0060] Specifically, if the specified evaluation method is the A-not-A sensory evaluation method, the form input data is verified, which specifically includes:

[0061] P1: Determine whether all form input data are positive integers. If they are all positive integers, execute P2; otherwise, the verification fails.

[0062] P2: Determine whether the number of samples determined as "A" in the A samples exceeds the total number of A samples, and determine whether the number of samples determined as "A" in the non-A samples exceeds the total number of A samples, that is, whether z i ≥n i . If neither exceeds (that is, z i <n i ), the verification passes; otherwise, the verification fails.

[0063] Among them, in S220, when automatically running the specified evaluation method, it includes:

[0064] S2201: Calculate the sensory difference intensity (d ′ value).

[0065] As an example, calculating the sensory difference intensity includes:

[0066] Q1: Construct a Probit regression model for the binomial distribution.

[0067] The form of the Probit regression model is: Φ -1 (P(Y = 1|x)) = β0 + β1x.

[0068] Among them, Φ -1 is the inverse cumulative distribution function of the standard normal distribution (i.e., the Probit link function); P(Y = 1|x) represents the probability that Y = 1 (indicating that the A sample or non-A sample is judged as "A") under the condition of the given independent variable x (representing the A sample or non-A sample); β0 and β1 are the regression coefficients of the Probit regression model, the former is the first intercept, and the latter is the first slope.

[0069] Q2: Obtain the regression coefficients of the Probit regression model based on maximum likelihood estimation, and the regression coefficients include the first intercept and the first slope.

[0070] As an example, obtaining the regression coefficients of the Probit regression model based on maximum likelihood estimation includes:

[0071] R1: Obtain the design matrix corresponding to the form input data based on the Probit regression model.

[0072] The following takes the data of two types of cigarettes shown in Table 1 as an example for illustration. Correspondingly, the form input data input through the user interface includes four groups of data.

[0073] Table 1

[0074]

[0075]

[0076] Among them, for the variable x, the A sample is denoted as 1, and the non-A sample is denoted as 0. In Table 1, x is denoted as [1, 1, 0, 0].

[0077] When calculating the design matrix, first calculate the probability P(Y = 1|x) = z i / n i .

[0078] In Table 1, all ni All are 20.

[0079] Thus, for x1,

[0080] for x2,

[0081] for x3,

[0082] for x4,

[0083] Secondly, calculate the Probit transformation value of each probability value. Specifically, use the inverse cumulative distribution function of the standard normal distribution (Φ -1 ) to transform each probability value:

[0084] For x1: Φ -1 (0.5) = 0;

[0085] For x2: Φ -1 (0.15) ≈ -1.04;

[0086] For x3: Φ -1 (0.5) = 0;

[0087] For x4: Φ -1 (0.85) ≈ 1.04.

[0088] Finally, use the Probit transformation value as the response variable y and the sample type x as the independent variable to construct a design matrix:

[0089] x y 1 0 1 -1.04 0 0 0 1.04

[0090] R2: Based on the design matrix, construct a linear regression function y i = β0 + β1x i + ε i . Among them, the intercept and slope of the linear regression function are the above-mentioned first intercept β0 and first slope β1, and ε i is the error term.

[0091] R3: Based on the maximum likelihood function, iterate on the first intercept and first slope. After the iteration is completed, obtain the predicted values of the first intercept and the first slope as the final first intercept and first slope.

[0092] For each independent variable x i , the likelihood function is: where Φ is the cumulative distribution function of the standard normal distribution and m is the number of independent variables.

[0093] In this application, the logarithm of the likelihood function is taken to simplify the calculation: If the specified assessment method is the A-non-A sensory assessment method, this assessment method belongs to binomial data. For binomial data, the complete log-likelihood function is written as where: η i = β0 + β1x i , is the z i corresponding to y i (In the examples of Table 1, z i is 10, 3, 10, 17).

[0094] As an example, using the numerical optimization method, the first intercept β0 and the first slope β1 are iteratively solved, as follows using the Newton–Raphson method.

[0095] Among them, when iterating the first intercept and the first slope based on the maximum likelihood function, in each round of iteration, the first score U0 of the first intercept β0 and the second score U1 of the first slope β1 are calculated based on the maximum likelihood function, and the Hessian matrix is constructed based on the maximum likelihood function; subsequently, the first intercept and the first slope are updated based on the initial first intercept and initial first slope, the first score, the second score, and the Hessian matrix of this round, and the updated first intercept and first slope are used as the initial first intercept and initial first slope of the next round. If the convergence condition is met, the first intercept and the first slope of the last round are used as the predicted value of the first intercept and the predicted value of the first slope, respectively.

[0096] Among them, the score function is: where j = 0 or 1, η i = β0 + β1x i , Φ is the cumulative distribution function of the standard normal distribution, is its density function. For the first intercept β0 (j = 0), then For the first slope β1 (j = 1), then

[0097] The score function is further illustrated below using the first round of iteration as an example.

[0098] Before the start of the first round (round number 0) of iteration, the first intercept β0 and the first slope β1 are first initialized. For example, β0 (0) = 0, β1 (0) = 0, and the superscript represents the round number. After initialization, the U0 and U1 of this round are calculated. Taking the example shown in Table 1, among them, for each xi , there is η i = β0 + β1x i = 0, Φ(η i ) = Φ(0) = 0.5, 0.3989. Therefore, For the first slope β1,

[0099] Continuing the above iterative steps, construct the Hessian matrix based on the maximum likelihood function. The elements H of the Hessian matrix jk The expression is where lnL is the value of the log-likelihood function, and β0 (l) is the initial first intercept at the l-th iteration, and β1 (l) is the initial first slope at the l-th iteration.

[0100] Among them, constructing the Hessian matrix includes:

[0101] S1: Calculate the first contribution matrix for sample type "A" and the second contribution matrix for sample type "non-A". The first contribution matrix and the second contribution matrix are two-row and two-column matrices. Determine the second intercept and the third intercept according to the first column of the first contribution matrix and the second contribution matrix respectively, and determine the second slope and the third slope according to the second column of the first contribution matrix and the second contribution matrix respectively.

[0102] Taking the above first-round iteration as an example, since β0 (0) = β1 (0) = 0, η i = 0, Φ(η i ) = Φ(0) = 0.5, for each independent variable x i , the contribution matrix is simplified to where X i is the two-column vector corresponding to the independent variable x i . The first column of this two-column vector is the intercept (always 1), denoted as X i0 = 1; the second column is the independent variable x i , that is, X i1 = X i .

[0103] For the example shown in Table 1, for x1 = 1 (i.e., sample A), X1 = [1, 1] T , For x2 = 1 (i.e., sample A), X2 = [1, 1] T , For x3 = 0 (i.e., non-A sample), X3 = [1, 0] T , For x4 = 0 (i.e., non-A sample), X4 = [1,0] T , In this example, the first contribution matrix for sample type "A" is Calculating the second intercept based on its first column gives -25.464 * 1 * 1 = -25.464; calculating the second slope based on its second column gives -25.464 * 1 * 1 = -25.464. The second contribution matrix for sample type "non-A" is Calculating the third intercept based on its first column gives -25.464 * 1 * 0 = 0; calculating the third slope based on its second column gives -25.464 * 0 * 0 = 0.

[0104] S2: Calculate the four element values of the Hessian matrix with two rows and two columns using the second intercept, third intercept, second slope, and third slope.

[0105] Specifically, according to the matrix positions, the four elements of the Hessian matrix are H 00 , H 01 , H 10 , H 11 .

[0106] H 00 is the sum of the second intercept and the third intercept, i.e., -25.464 + 0 = -25.464;

[0107] H 01 is the sum of the second intercept and the third slope, i.e., -25.464 + 0 = -25.464;

[0108] H 10 is the sum of the second slope and the third intercept, i.e., -25.464 + 0 = -25.464;

[0109] H 11 is the sum of the second slope and the third slope, i.e., -25.464 + 0 = -25.464.

[0110] Therefore, the initially constructed Hessian matrix is

[0111] Continuing the above iterative steps, in the first round of iteration, after obtaining the first score, second score, and the Hessian matrix, update the first intercept and the first slope according to the following formula:

[0112]

[0113] Calculate the new first intercept and first slope through the above formula, and then use the updated first intercept and first slope as the initial first intercept and initial first slope for the next round of iteration. If the convergence condition is met (that is, the change values of the first intercept and first slope in two adjacent iterations are less than the threshold), then use the first intercept and first slope of the last round as the predicted values of the first intercept and the predicted value of the first slope

[0114] In the example of Table 1 above,

[0115] Q3: Take the opposite of the first slope β1 as the sensory difference intensity, that is, d' = -β1.

[0116] In the example of Table 1 above, d′ = 1.036.

[0117] S2202: Calculate the significance of the difference between samples (p-value), and set the significance level threshold to α = 0.05.

[0118] As an example, use Fisher's exact test to calculate the p-value. Specifically, it includes the following steps:

[0119] U1: Construct a contingency table with sample data, and the table format is as follows:

[0120] Judged as "A" Judged as "not A" Total Sample A a c <![CDATA[n1]]> Sample not A b d <![CDATA[n2]]> Total n = a + b c + d <![CDATA[N = n1 + n2]]>

[0121] Among them, the number of samples in sample A determined as "A" is a; the total number of sample A is n1, c = n1 - a; the number of samples in non-A samples determined as "A" is b; the total number of non-A samples is n2, d = n2 - b; N = n1 + n2.

[0122] U2: Calculate the first probability that the number of samples in sample A determined as "A" is k ≥ a, and take this first probability as the p-value.

[0123] Specifically, while keeping the row and column totals of the above table unchanged, list all contingency tables with k ≥ a, and calculate the probability corresponding to each k.

[0124] For the current row and column totals, the possible table forms are:

[0125] Judged as "A" Judged as "not A" Sample A k a + c - k Sample not A n - k k + d - a

[0126] Among them, the value range of k is an integer that makes all cells non-negative, and n is the total number selected in the sample (in this example, it is the total number of samples in "A" samples determined as "A" and "non-A" samples determined as "A")

[0127] As an example, the hypergeometric distribution is used to calculate the probability corresponding to each k, and the difference between the observed distribution and the expected distribution in the contingency table is calculated:

[0128]

[0129] where represents the combination number, that is, the number of ways to select k elements from n1 elements; represents the number of ways to select n - k elements from n2 elements; represents the number of ways to select n elements from N elements.

[0130] For example, as an instance, a = 10, b = 3, n = 13, n1 = 20, n2 = 20, N = 40, then Therefore, P(X = 10) = 184756×1140 / 8653443109 ≈ 0.0245.

[0131] Calculate the sum of the probabilities of all possible k values, that is, the p - value of the Fisher's exact test, that is, p = P(X≥a). In the above instance, P(X≥10) = P(X = 10)+P(X = 11)+P(X = 12)+…+P(X = 13) = 0.0203712.

[0132] U3: Compare the p - value with the significance level threshold α = 0.05. If p < 0.05, it is considered that there is a significant difference between the samples.

[0133] In the above instance, since p < 0.05, it indicates that there is a significant difference between sample A and non - A samples at the significance level α = 0.05, that is, the sensory difference between the two cigarettes in this instance reaches a significant level.

[0134] After the operation is completed, the sensory difference intensity and the significance of the difference between samples are output. As an example, the user interface outputs the following content: formatted text results (including d' value, p - value); automatically generates conclusions such as "p < 0.05, there is a significant difference between samples", which is convenient for users to quickly interpret the analysis results. This module combines a graphical interface with text output, enhancing the comprehensibility and intuitiveness of the results and improving the user experience.

[0135] Preferably, the present application also has a data export function: the system supports exporting the analysis results in Excel and CSV formats, which is convenient for subsequent data analysis and report generation.

[0136] The beneficial technical effects of the present application are as follows:

[0137] 1. Convenient operation: The parameter input process is simplified through a graphical interface, allowing users to quickly complete the A-not-A sensory evaluation without programming knowledge, significantly reducing the technical threshold.

[0138] 2. High efficiency and accuracy: Built-in optimized calculation algorithms support fast calculation responses in milliseconds.

[0139] 3. Clear results: The calculation results are presented through intuitive text combinations, helping users better understand the sensory differences between samples and improving the accuracy of decision-making.

[0140] 4. Strong scalability: The modular design enables the system to have good scalability. In the future, more sensory evaluation methods can be integrated according to requirements to support applications in different fields.

[0141] Although some specific embodiments of the present application have been described in detail through examples, those skilled in the art should understand that the above examples are only for illustration and not for limiting the scope of the present application. Those skilled in the art should understand that the above embodiments can be modified without departing from the scope and spirit of the present application. The scope of the present application is defined by the appended claims.

Claims

1. A sensory evaluation method, characterized in that: include: In response to receiving a designated instruction of an evaluation method from a user interaction interface, outputting a form input interface corresponding to the designated evaluation method; In response to receiving form input data from the form input interface, automatically running the specified assessment method; The operation result of the specified evaluation method is output through the user interaction interface.

2. The sensory evaluation method according to claim 1, characterized in that: After receiving the form input data from the form input interface, the form input data is verified; if the verification passes, the specified evaluation method is automatically run.

3. The sensory evaluation method according to claim 1, characterized in that: If the designated evaluation method is an A-non-A sensory evaluation method, the automatic operation of the designated evaluation method includes: Calculate sensory difference intensity; Calculate the significance of differences between samples.

4. The sensory evaluation method according to claim 2, characterized in that: If the specified evaluation method is the A-non-A sensory evaluation method, the form input data is verified, specifically including: Determine whether all form input data is a positive integer; If both are true, then determine whether the number of samples judged as "A" in the A samples exceeds the total number of A samples, and determine whether the number of samples judged as "A" in the non-A samples exceeds the total number of A samples; If neither exceeds, the verification passes.

5. The sensory evaluation method according to claim 3, characterized in that: Calculate sensory difference intensity, including: Construct a Probit regression model for binomial distribution; Obtaining the regression coefficient of the Probit regression model based on maximum likelihood estimation, wherein the regression coefficient includes a first intercept and a first slope; The inverse of the first slope is the sensory difference intensity.

6. The sensory evaluation method according to claim 5, characterized in that: The regression coefficients of the Probit regression model are obtained based on maximum likelihood estimation, including: Obtaining a design matrix corresponding to the form input data based on a Probit regression model; constructing a linear regression function based on the design matrix, wherein the intercept and slope of the linear regression function are the first intercept and the first slope; The first intercept and the first slope are iterated based on the maximum likelihood function, and a predicted value of the first intercept and a predicted value of the first slope are obtained after the iteration is completed.

7. The sensory evaluation method according to claim 6, characterized in that: Iterating the first intercept and the first slope based on a maximum likelihood function includes: In each round of iteration, a first score of the first intercept and a second score of the first slope are calculated based on the maximum likelihood function, and a Hessian matrix is ​​constructed based on the maximum likelihood function; then, the first intercept and the first slope are updated based on the initial first intercept and the initial first slope of this round, the first score, the second score, and the Hessian matrix, and the updated first intercept and the first slope are used as the initial first intercept and the initial first slope of the next round; If the convergence condition is met, the first intercept and the first slope of the last round are used as the predicted value of the first intercept and the predicted value of the first slope, respectively.

8. The sensory evaluation method according to claim 7, characterized in that: Construct the Hessian matrix, including: Calculate a first contribution matrix for sample type "A" and a second contribution matrix for sample type "non-A", wherein the first contribution matrix and the second contribution matrix are two-row and two-column matrices, and determine a second intercept and a third intercept according to the first column of the first contribution matrix and the second contribution matrix, and determine a second slope and a third slope according to the second column of the first contribution matrix and the second contribution matrix; The four element values ​​of the Hessian matrix with two rows and two columns are calculated using the second intercept, the third intercept, the second slope and the third slope.

9. A sensory evaluation system, characterized in that: It comprises a user interaction interface and a background processing module, and the background processing module is used to execute the sensory evaluation method according to claims 1-7.

10. The sensory evaluation system according to claim 8, characterized in that: The background processing module is encapsulated as a dynamic link library.

Images