Sensory evaluation method and sensory evaluation system

Abstract

To provide a sensory evaluation method and a sensory evaluation system by a pair comparison method with less burden on an evaluator and high accuracy.SOLUTION: A sensory evaluation method and a sensory evaluation system in which a pair of samples are extracted from a plurality of samples and evaluated by an evaluator, wherein the pair of samples of N * (N - 1) / 2 combinations extracted from a total number N of the samples are evaluated by a plurality of evaluators in charge of the same number of combinations without including an order effect, A structural formula showing the evaluation result of an evaluator (k) at the time of comparing a sample (I) with a sample (j) is shown by Xijk = (α I - α j) + ε ijk when the average liking degree of each sample is defined as α and the error of each average liking degree is defined as ε ijk, and the average liking degree α I of the sample (I) is calculated by α I = (1 / N) * Xi.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to a sensory evaluation method and a sensory evaluation system for a vehicle, and to an evaluation method using a paired comparison method. [Background technology]

[0002] Sensory evaluation using human senses is sometimes used as a method for evaluating things that are difficult to quantify through measurement, such as the ride comfort and sound quality of a vehicle. For example, Patent Document 1 describes a sound quality evaluation system in which an evaluator (subject) listens to and evaluates the sound quality of a vehicle. In this system, the evaluator compares and scores a pair of sample sounds (paired comparison method), and multiple evaluators share the work of comparing all combinations of the sample sounds.

[0003] Then, statistical processing is performed on the evaluation results of each set to obtain an evaluation of the test sound. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2022-74784 Summary of the Invention [Problem to be solved by the invention]

[0005] In Patent Document 1, the task is assigned according to the ability, preference, etc. of the evaluator. Therefore, there is a problem that the assignment of the task is time-consuming, and furthermore, the calculation load increases because statistical processing is performed using a matrix of the evaluation results. The present invention has been made to solve such problems, and an object of the present invention is to provide a sensory evaluation method and a sensory evaluation system using the paired comparison method that reduces the burden on the evaluator and enables easy and accurate evaluation. [Means for solving the problem]

[0006] In order to achieve the above-mentioned object, the sensory evaluation method of the present invention is a sensory evaluation method in which a pair of samples is selected from a plurality of samples and evaluated by an evaluator, wherein N×(N-1) / 2 combinations of pairs of samples are selected from a total number N of the samples, and the plurality of evaluators are each responsible for evaluating the same number of combinations without taking into account order effects, and the structural formula showing the evaluation result of evaluator k when comparing sample i and sample j is expressed as Xijk = (αi-αj)+εijk, where α is the average preference score for each sample and εijk is the error in each average preference score.

[0007] The sensory evaluation system of the present invention is a sensory evaluation system in which a pair of samples is selected from a plurality of samples and evaluated by an evaluator, and includes an input unit for inputting the evaluation results of N×(N-1) / 2 combinations of pairs of samples selected from a total of N samples, with the plurality of evaluators dividing the evaluation results into equal numbers of combinations without including order effects; a memory unit for storing the evaluation results input from the input unit; and a calculation unit for calculating an average preference score for each sample based on the evaluation results stored in the memory unit, wherein the calculation unit calculates the average preference score αi for sample i by αi=(1 / N)×Xi..., where Xi... is the sum of the evaluation results including sample i. [Effects of the Invention]

[0008] According to the sensory evaluation method and the sensory evaluation system of the present invention, the evaluator evaluates the combination of a pair of samples without taking into account the order effect, and thus the number of combinations to be evaluated can be reduced to half. Furthermore, by having multiple people share the task of evaluating these paired combinations, the number of evaluations per person can be reduced, thereby reducing the burden on the evaluators.

[0009] Furthermore, by having multiple evaluators divide the evaluation among the same number of combinations without including the order effect, the average preference level can be easily calculated without having to consider individual differences in preference level, and each sample can be evaluated easily and accurately. [Brief explanation of the drawings]

[0010] [Figure 1] FIG. 10 is a diagram showing the allocation of tasks among evaluators in the evaluation method according to one embodiment of the present invention when the total number of samples is an even number. [Figure 2] FIG. 10 is a diagram showing the allocation of tasks among evaluators in the evaluation method according to one embodiment of the present invention when the total number of samples is an odd number. [Figure 3] 1 is a configuration diagram of an evaluation system according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0011] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. One embodiment of the present invention is a sensory evaluation method and a sensory evaluation system in which a human (evaluator) evaluates an evaluation object that is difficult to evaluate by measurement results, such as the sound of a vehicle. The evaluation method (sensory evaluation method) of this embodiment is a paired comparison method in which a pair of samples is extracted from a large number (three or more) of evaluation objects (samples) and evaluated.

[0012] In the evaluation, the evaluator compares sample i and sample j and assigns a score on multiple levels to indicate which is better or how much better each is. For example, when comparing sample i and sample j, Sample i is much better than sample j...+2 points Sample i is slightly better than sample j...+1 point No difference between sample i and sample j: 0 points Sample j is slightly better than sample i... -1 point Sample j is much better than sample i... -2 points The evaluator will then decide on one of five possible scores.

[0013] When evaluating a total of N samples, the combination of pairs of samples to be evaluated is NC2 (= N × (N - 1) / 2) ways are possible. In this embodiment, the effect of the order in which samples are evaluated is not taken into consideration. For example, if the samples are sounds, listening to sample i first and then sample j will have the same absolute value as listening to sample j first and then sample i, with only the signs (+ -) of the scores reversed.

[0014] Furthermore, in this embodiment, all the combinations of pairs of samples are evaluated by a plurality of evaluators. If the total number of samples N (total number of samples N) is an even number, there are N × (N-1) / 2 possible combinations of samples, which are divided among M = N / 2 evaluators, with each evaluator performing N-1 evaluations.

[0015] If the total number of samples N is odd, the N × (N-1) / 2 combinations of samples are divided among the number of evaluators M = (N-1) / 2, with each evaluator conducting N evaluations. Figure 1 is a table showing an example of how to allocate samples when the total number of samples N is an even number. Figure 2 is a table showing an example of how to allocate samples when the total number of samples N is an odd number. 1 and 2, Xijk shown in the tables indicates the evaluation result when evaluator k compared samples i and j. -Xijk in the tables is the value obtained by reversing the sign (+-) of Xijk.

[0016] If the total number of samples N is an even number, for example, if the total number of samples N is 6, the evaluations will be conducted by three evaluators (total number of samples N = 6, number of evaluators M = 3). As shown in Figure 1, out of the 6 × 5 / 2 = 15 combinations without considering the order effect, each evaluator will evaluate 15 / 3 = 5 combinations. As shown in Figure 1(A), evaluator A evaluates X121, X231, X341, X451, and X561 where k = 1. Furthermore, the signs (+ / -) of these evaluation results (scores) are reversed to give evaluations of -X121, -X231, -X341, -X451, and -X561.

[0017] Evaluator B evaluates X132, X242, X352, X462, and X162, where k=2. In addition, the signs (+ / -) of these evaluation results (scores) are reversed to give the evaluations -X132, -X242, -X352, -X462, and -X162. Evaluator C evaluates X143, X253, X363, X153, and X263 where k=3. In addition, the signs (+ / -) of these evaluation results (scores) are reversed to give evaluations of -X143, -X253, -X363, -X153, and -X263.

[0018] As a result, as shown in Figure 1(D), a total of six samples (N = 6) are divided among three evaluators, with each evaluator performing five evaluations, for a total of 15 evaluations, resulting in the acquisition of all 30 evaluations. Furthermore, the evaluations are distributed to prevent one evaluator from evaluating the same samples in a biased manner. If the total number of samples N is odd, for example, if the total number of samples N is 7, the evaluations will be conducted by three evaluators (N=7, M=3). As shown in Figure 2, out of the 7 × 6 / 2 = 21 combinations without considering the order effect, each evaluator will evaluate 21 / 3 = 7 combinations.

[0019] As shown in Figure 2(A), evaluator A evaluates X121, X231, X341, X451, X561, X671, and X171 where k = 1. Furthermore, the signs (+ / -) of these evaluation results (scores) are reversed to give evaluations of -X121, -X231, -X341, -X451, -X561, -X671, and -X171. Evaluator B evaluates X132, X242, X352, X462, X572, X672, X162, and X272, where k=2. He also reverses the signs (+ / -) of these evaluation results (scores) to obtain the evaluations -X132, -X242, -X352, -X462, -X572, -X672, -X162, and -X272.

[0020] Evaluator C evaluates X143, X253, X363, X473, X153, X263, and X373, where k=3. In addition, the signs (+ / -) of these evaluation results (scores) are reversed to give the evaluations -X143, -X253, -X363, -X473, -X153, -X263, and -X373. As a result, as shown in Figure 2(D), a total of N = 7 samples are divided among three evaluators, with each evaluator performing 7 evaluations, for a total of 21 evaluations, resulting in a total of 42 evaluations. In addition, the evaluations are distributed to prevent one evaluator from evaluating the same samples in a biased manner.

[0021] Next, the structural formulas of the evaluation results will be explained. First, the structural formula in the known method is shown below. In the following structural formula, Xijk is the result of the kth evaluator when comparing i and j, αi is the average preference, αik is the individual difference in preference, δi is the order effect, and εijk is the error in each average preference. 1) Ura's modified method: Each rater evaluates all combinations (including those with different orders) taking into account the order effect.

[0022] Xijk = (αi - αj)+(αik - αjk) )+(δ + δi)+ γi j+εi ···(1) 2) Nakaya's modified method: Each rater evaluates all combinations without considering the order effect. Xijk = (αi - αj)+(αik - αjk)+γi j+εi ···(2) 3) Cyclic paired comparison method: Multiple raters divide the work and evaluate all combinations, taking into account the order effect.

[0023] Xijk = (αi - αj)+(αik - αjk)+(δ + δi)+εijk (3) JPEG2026031113000002.jpg25132In contrast, in this embodiment, multiple samples are divided among multiple evaluators, and all pair combinations are evaluated without taking into account the order effect, with each evaluator performing the evaluation the same number of times, resulting in a structural formula shown in formula (4) below.

[0024] Xijk = (αi - αj)+εijk ···(4) That is, this embodiment is a general solution for one person of Nakaya's modified method. As a result, the average preference score αi can be calculated by the following equation (5). αi=(1 / N)×Xi (5) It should be noted that Xi.. is the total evaluation value when sample i=1 is used, and when i=1, X1..=X12.+X13.+X14.+X15.+X16. . . .

[0025] For example, in the case of Figure 1(d), X1..=X121+X132+X143+X153+X162=-{(-X121)+(-X132)+(-X143)+(-X153)+(-X162)} X2..=(-X121)+X231+X242+X253+X263=-{X121+(-X231)+(-X242)+(-X253)+(-X263)} X3..=(-X132)+(-X231)+X341+X352+X363=-{X132+X231+(-X341)+(-X352)+(-X363)} X4..=(-X143)+(-X242)+(-X341)+X451+X462=-{X143+X242+X341+(-X451)+(-X462)} X5..=(-X153)+(-X253)+(-X352)+(-X451)+X561=-{X153+X253+X352+X451+(-X561)} X6..=(-X162)+(-X263)+(-X363)+(-X462)+(-X561)=-{X162+X263+X363+X462+X561} Therefore, the average preference score αi for each sample can be easily calculated using the total of these evaluation values ​​and equation (5).

[0026] As described above, the evaluation method of this embodiment is a sensory evaluation method in which an evaluator selects a pair of samples from a plurality of samples and evaluates them. By evaluating N×(N−1) / 2 combinations of pairs of samples selected from a total of N samples without taking into account the order effect, the number of evaluations can be reduced to 1 / 2. Furthermore, by having multiple people share the task of evaluating these paired combinations, the number of evaluations per person can be reduced, thereby reducing the burden on the evaluators.

[0027] Furthermore, by having multiple evaluators evaluate the same number of combinations and using a data structure of Xijk = (αi - αj) + εijk, the average preference level αi can be easily calculated without having to consider individual differences in preference levels, making it easy to evaluate each sample. If the total number of samples, N, is an even number, N × (N-1) / 2 combinations of pairs of samples can be divided among N / 2 raters to evaluate them, so that the number of evaluations per person is N-1. This makes it possible to easily calculate the average preference score, αi, without having to consider individual differences in preference scores.

[0028] Furthermore, if the total number of samples N is odd, N × (N-1) / 2 combinations of pairs of samples can be divided among (N-1) / 2 evaluators to evaluate them, so that the number of evaluations per person is N. This makes it possible to easily calculate the average preference score αi without having to consider individual differences in preference scores, as described above. FIG. 3 is a diagram showing the configuration of an example of an evaluation system for carrying out the sensory evaluation method of the present invention.

[0029] The evaluation system 1 of this embodiment includes an input unit 2, a control unit 3, and an output unit 4. The input unit 2 is a device for inputting the results of the evaluation of a pair of samples by an evaluator, and is a device capable of inputting the above evaluation points, such as a keyboard or switch. The input unit 2 can also input the total number of samples N. The control unit 3 is configured with built-in input / output devices, storage devices (ROM, RAM, non-volatile RAM, etc.), a CPU, etc., and has a storage unit 5 and a calculation unit 6.

[0030] The control unit 3 determines the number of evaluators M based on the total number of samples N input from the input unit 2 before the evaluation, and also determines the combinations of pairs of samples and the evaluators.After the evaluators input their evaluation results into the input unit 2, the control unit 3 stores the evaluation results input from the input unit 2 in a matrix in the memory unit 5 for each combination of samples. The control unit 3 may sequentially display and specify the combinations of samples to be evaluated and the evaluators on the output unit 4 based on the total number of samples N previously input into the input unit 2, and have the evaluators perform the evaluations. Alternatively, the evaluators may evaluate any combinations, input the evaluated pairs of samples, and then input the evaluation results. However, to prevent multiple evaluations of each combination, the output unit 4, such as a display device, may notify combinations for which evaluation results have not been input.

[0031] After all the combinations have been evaluated, the calculation unit 6 reads out the evaluation results from the storage unit 5 and calculates the preference level of each sample using the above formula (5). The output unit 4 is, for example, a display panel, and provides the preference level for each sample calculated by the calculation unit 6 by display or the like. The evaluation system 1 configured as above makes it possible to carry out the sensory evaluation method described above.

[0032] The present invention is not limited to the above-described embodiments. For example, in the above evaluation method, the types and total number of evaluation targets may be changed as appropriate. Also, various configurations of the evaluation system 1 may be changed as appropriate. [Explanation of symbols]

[0033] 1. Rating System 2 Input section 5 Storage section 6 Arithmetic section

Claims

1. A sensory evaluation method in which a pair of samples is extracted from a plurality of samples and evaluated by an evaluator, A pair of N × (N-1) / 2 combinations of samples extracted from the total number N of samples are evaluated by multiple evaluators, each of whom is responsible for the same number of combinations, without taking into account the order effect; The structural formula showing the evaluation results of evaluator k when comparing sample i and sample j is expressed as Xijk = (αi-αj)+εijk, where α is the average preference for each sample and εijk is the error in each average preference. A sensory evaluation method characterized by:

2. When the total number N of the samples is an even number, N×(N−1) / 2 combinations of pairs of the samples are divided among N / 2 evaluators to be evaluated. The sensory evaluation method according to claim 1 .

3. When the total number N of the samples is odd, N×(N−1) / 2 combinations of pairs of the samples are divided among (N−1) / 2 evaluators to evaluate them. The sensory evaluation method according to claim 1 .

4. The evaluation involves determining which of the pair of samples is preferable and assigning each sample a score on a scale of one to another. The sensory evaluation method according to claim 1 .

5. When one sample and the other sample are evaluated in the order of N x (N-1) / 2 combinations of pairs of samples extracted from the total number N of samples, the evaluation of the other sample and the other sample in the order of one sample and the other sample is made the score with the sign reversed from the result of evaluation of the one sample and the other sample in the order of one sample and the other sample. The sensory evaluation method according to claim 4 .

6. A sensory evaluation system in which a pair of samples is extracted from a plurality of samples and evaluated by an evaluator, an input unit for inputting the results of evaluation of N×(N−1) / 2 combinations of pairs of samples extracted from a total number N of samples, with multiple evaluators dividing the evaluation into the same number of combinations without including order effect; a storage unit that stores the evaluation results input from the input unit; a calculation unit that calculates an average preference score for each sample based on the evaluation results stored in the storage unit, The calculation unit calculates the average preference score αi of sample i by αi = (1 / N) × Xi, where Xi is the sum of the evaluation results including sample i. A sensory evaluation system characterized by:

Images