A low-salt food palatability quantitative evaluation method and system based on multi-modal brain neural signals

Abstract

This invention discloses a method and system for quantitatively evaluating the palatability of low-salt foods based on multimodal brain neural signals. The method involves acquiring a sample of the low-salt food to be evaluated and a reference sample, presenting the samples according to a preset procedure, and simultaneously acquiring electroencephalogram (EEG) signals and functional near-infrared cerebral oxygenation response (FIR) signals. After preprocessing and feature extraction, EEG, event-related potential (ERP), and FIR signal characteristics are obtained and fused to form a feature vector. This vector is then input into a palatability quantitative evaluation model to obtain a comprehensive palatability score and flavor dimension sub-scores. Combining the actual salt reduction rate and palatability retention rate, the method outputs a palatability grade and a matching degree between the salt reduction rate and palatability. This invention reduces reliance on single subjective sensory evaluations, improves the objectivity and repeatability of the evaluation, and is applicable to low-salt food research and development, formula screening, and quality control.

Description

Technical Field

[0001] This invention belongs to the fields of food sensory evaluation, intelligent food detection, neuroscience and low-salt food development technology, specifically involving a method and system for quantitative evaluation of the palatability of low-salt foods based on multimodal brain neural signals. Background Technology

[0002] Excessive sodium intake increases the risk of chronic diseases such as hypertension and cardiovascular disease. To meet the health needs of residents, the food industry continues to promote the development of low-sodium and salt-reduced processed foods. Salt not only provides the basic saltiness in food systems but also affects umami perception, aroma release, textural properties, and overall flavor harmony. After reducing salt content, meat products, compound seasonings, prepared dishes, fermented foods, and snack foods are prone to problems such as weakened saltiness, insufficient umami, bland flavor, or unbalanced aroma, thus reducing the overall acceptability of the product. Therefore, how to maintain good palatability while effectively reducing salt content has become a crucial technical problem that needs to be solved in the research and development and industrial application of low-sodium food formulations.

[0003] Currently, the palatability evaluation of low-sodium foods still mainly relies on artificial sensory evaluation, with common methods including saltiness intensity scoring, overall preference scoring, ranking tests, and descriptive sensory analysis. While artificial sensory evaluation is intuitive, it still has the following limitations: First, the evaluation results are easily affected by factors such as the evaluator's dietary preferences, evaluation experience, emotional state at the time, and testing environment, resulting in limited objectivity, repeatability, and stability of the evaluation results. Second, the evaluation dimensions are relatively singular. Some evaluation schemes mainly focus on measuring the intensity of saltiness, making it difficult to simultaneously quantify the multi-dimensional perceptual characteristics of low-salt foods, such as saltiness, umami, spiciness, aroma, and overall pleasantness, which is not conducive to the comprehensive palatability judgment. Third, the verbal ratings or questionnaire feedback of the subjects may not be completely consistent with their neural preference responses, and traditional sensory evaluations are difficult to directly reflect the differences in neural responses related to food flavor acceptance. Fourth, existing evaluation results are usually expressed in terms of individual sensory scores or overall preference, which makes it difficult to further quantify the matching relationship between actual salt reduction rate and product palatability. This provides limited support for low-salt formula screening, product flavor grading, and production quality control.

[0004] Electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) are techniques for detecting brain signals that can acquire the brain electrical activity and cerebral blood oxygenation response generated when subjects are stimulated by taste and smell. Event-related potentials (ERPs) can be extracted from continuous EEG signals based on event markers. These brain responses can be used to analyze neural activity related to attention, stimulus processing, and preference evaluation, providing a new technical approach to improve the objectivity of food palatability evaluation.

[0005] However, existing research mainly focuses on single-taste stimuli or single-modality brain neural signal analysis, lacking standardized evaluation schemes applicable to various low-salt foods such as meat products, seasonings, and prepared dishes. Furthermore, current technologies lack a quantitative palatability evaluation system that integrates brain electrical neural response characteristics, event-related potential characteristics, and cerebral blood oxygenation response characteristics with evaluation indicators such as saltiness, umami, spiciness, aroma, and overall pleasantness. This makes it difficult to simultaneously output a comprehensive palatability score, palatability grade, and the degree of matching between salt reduction rate and palatability.

[0006] In summary, there is a need to provide a method and system for quantitatively evaluating the palatability of low-salt foods based on multimodal brain neural signals, so as to improve the objectivity and repeatability of the palatability evaluation of low-salt foods, realize the quantitative output of comprehensive palatability score, grading determination and matching degree between salt reduction rate and palatability, and provide technical support for the screening, optimization and quality control of low-salt food formulas. Summary of the Invention

[0007] The purpose of this invention is to provide a method and system for quantitatively evaluating the palatability of low-salt foods based on multimodal brain neural signals, in order to solve the problems of existing sensory evaluation of low-salt foods being easily affected by subjective factors, lacking repeatability, having relatively single evaluation dimensions, being unable to reflect the difference between subjective evaluation and neural preferences, and being unable to quantify the matching relationship between salt reduction rate and palatability.

[0008] This invention synchronously collects EEG signals and functional near-infrared cerebral oxygenation response signals of subjects under low-salt food stimulation, extracts neural response features related to saltiness, umami, oral stimulation, aroma, pleasure, and overall acceptability, and constructs a quantitative evaluation model for the palatability of low-salt foods. This aims to improve the objectivity and repeatability of the palatability evaluation of low-salt foods, and to assist in palatability scoring, grade determination, and formula optimization.

[0009] To achieve the above objectives, the present invention provides the following technical solution: First, a method for quantitatively evaluating the palatability of low-salt foods based on multimodal brain neural signals is provided, including the following steps: S1. Sample Information Acquisition and Standardized Stimulus Presentation Obtain the low-salt food samples to be evaluated and their corresponding reference samples. The low-salt food samples to be evaluated include, but are not limited to, low-salt seasonings, low-salt meat products, low-salt salt water products, low-salt convenience foods, low-salt snack foods, low-salt pre-cooked dishes, low-salt fermented foods, or low-salt compound seasoning foods; the reference samples can be food samples with the same or similar food matrix as the low-salt food samples to be evaluated, standard salt content food samples, conventional formula food samples, commercially available food samples of the same category, or benchmark samples selected from preset salt reduction rate gradient samples.

[0010] Record the basic information of the low-salt food sample to be evaluated and the reference sample. The basic information includes one or more of the following: food category, salt content or sodium content, formula information, processing method, storage conditions and target consumer group. Record the target salt reduction rate of the low-salt food sample to be evaluated.

[0011] The low-salt food sample to be evaluated and the reference sample were presented to the subjects according to a pre-set sensory stimulation procedure. The samples were numbered, randomized, and blinded, and the sample temperature, sample mass or volume, single intake, stimulation duration, stimulation interval, and rinsing or taste recovery time were controlled. After completing baseline signal acquisition in a quiet environment, the subjects then came into contact with, smelled, or ingested the low-salt food sample to be evaluated and the reference sample in a pre-set sequence.

[0012] For liquid or semi-solid samples, a quantitative oral stimulation method can be used; for solid samples, a standardized chewing time and pre-swallowing evaluation method can be used; for samples with obvious aroma characteristics, olfactory stimulation stage and oral stimulation stage can be set separately.

[0013] S2, Synchronous Acquisition of Multimodal Brain Nerve Signals During the presentation of the sample stimulus, the subject's electroencephalogram (EEG) signals and functional near-infrared cerebral oxygenation response signals were simultaneously acquired, and event tagging was used to achieve temporal synchronization between the stimulus events and different modalities of brain nerve signals. The functional near-infrared cerebral oxygenation response signals include oxyhemoglobin concentration change signals and deoxyhemoglobin concentration change signals.

[0014] The event-related potential features are time-domain neural response features obtained by segmenting continuous EEG signals based on event markers, performing baseline correction, removing abnormal trials, and averaging. They are not considered as independently acquired brain neural signal modalities.

[0015] The electroencephalogram (EEG) signals are used to characterize neural oscillations during food taste, smell, and related cognitive, emotional, and preference processing; the event-related potential (ERP) characteristics are used to reflect changes in neural activity related to attention allocation, stimulus recognition, and evaluation processing; and the functional near-infrared cerebral oxygenation response signals are used to characterize blood oxygen dynamics changes in measurable superficial cortical regions such as the prefrontal cortex.

[0016] During signal acquisition, key time points such as sample presentation, olfaction start, oral ingestion start, chewing start, swallowing, scoring start, or result submission are recorded synchronously through event markers.

[0017] S3. Signal Preprocessing and Feature Extraction The collected electroencephalogram (EEG) signals and functional near-infrared cerebral oxygenation response signals were preprocessed, and neural response features related to the palatability of low-salt foods were extracted from the effective brain neural response signals after preprocessing.

[0018] The EEG signal preprocessing includes one or more of the following: bandpass filtering, notch filtering, bad conduction identification and processing, independent component analysis, blink artifact removal, electromyography artifact removal, and baseline correction. Based on event markers, the preprocessed continuous EEG signals are segmented, baseline corrected, abnormal trial eliminations are performed, and averaged to obtain event-related potential responses.

[0019] The functional near-infrared brain oxygenation response signal preprocessing includes one or more of motion artifact correction, low-pass filtering, baseline correction, and blood oxygen concentration change calculation to obtain oxyhemoglobin concentration change signals and deoxyhemoglobin concentration change signals.

[0020] Brain electrical response features, event-related potential features, and cerebral blood oxygenation response features were extracted from the preprocessed effective brain neural response signals. Among them: (1) The EEG neural response features include one or more of the following: EEG power spectrum features, EEG time-frequency features, EEG coherence features, EEG complexity features, frontal lobe asymmetry features, and brain region functional connectivity features; (2) The event-related potential characteristics include the peak amplitude of the event-related potential, the peak latency, the average amplitude within a preset time window, and one or more of the amplitude and latency of the P300 component or the late positive potential component; (3) The brain oxygenation response characteristics include one or more of the following: peak oxygenation response, area of ​​oxygenation response, slope of oxygenation response, and brain region connectivity characteristics.

[0021] Based on the neural response characteristics, a feature set related to different perceptual dimensions of low-salt foods is constructed, wherein the perceptual dimension features include: ① Characteristics related to salty taste perception: including changes in brain electrical power, event-related potential changes, and changes in cerebral blood oxygenation response in measurable cortical areas such as the prefrontal cortex after taste stimulation; ② Features related to umami perception: including changes in EEG oscillations related to umami stimuli, changes in event-related potentials, and changes in cerebral blood oxygenation response in areas such as the prefrontal cortex; ③ Oral stimulation-related features: including neural response features related to the intensity of stimuli, attention allocation, and somatic sensory processing of spiciness, numbing sensation, burning sensation, cooling sensation, or other oral stimuli; ④ Aroma perception-related features: including changes in EEG frequency band power during the olfactory stage, changes in event-related potentials, and prefrontal cortex response features related to aroma perception and aroma pleasure; ⑤ Characteristics related to pleasure: including frontal lobe asymmetry, prefrontal cortex oxygenation response related to pleasure evaluation, and changes in P300 or late positive potential (LPP) components; ⑥ Overall acceptability-related characteristics: including EEG complexity, brain region functional connectivity, collaborative response of different brain regions, and cross-modal integrated response characteristics.

[0022] For different categories of low-salt foods, one or more of the aforementioned perceptual dimension features are selected based on their flavor attributes; for samples that do not have obvious oral irritation or aroma characteristics, the corresponding perceptual dimension features may not be extracted.

[0023] S4, Multimodal Feature Fusion Multimodal feature fusion was performed on the EEG neural response features, event-related potential features, and cerebral blood oxygenation response features obtained in step S3. Missing values ​​were processed, standardized, and feature selection was performed on different types of neural response features. The processed features were then concatenated to obtain the fused feature vectors corresponding to the low-salt food sample to be evaluated and the reference sample, respectively.

[0024] S5. Palatability Model Construction and Scoring Output Based on the fused feature vectors of the low-salt food samples to be evaluated and the reference samples obtained in step S4, and combined with one or more of the subjects' subjective sensory scores, overall preference scores, purchase intentions and preference selection results, a quantitative evaluation model for the palatability of low-salt foods is constructed.

[0025] During model construction, the fused feature vector of the training samples is used as the model input. The standardized overall preference score, or the comprehensive sensory score obtained from subjective scores of multiple sensory dimensions, is used as the regression training label for the palatability comprehensive score. The subjective evaluation results of sensory dimensions such as saltiness, umami, oral stimulation, aroma, and pleasure are used as the training labels for the corresponding flavor dimension sub-scores. The preset palatability level or preference selection results are used as the classification training labels.

[0026] The sample data was divided into training, validation and test sets according to the subjects to avoid data from the same subject being included in both the training and test sets at the same time. The palatability quantification evaluation model was determined through cross-validation, parameter optimization and model performance evaluation.

[0027] The palatability quantification evaluation model is a regression model, or a combination of a regression model and a classification model. The regression model may be one or more of partial least squares regression, support vector regression, random forest regression, and gradient boosting tree regression; the classification model may be one or more of support vector classification, random forest classification, and gradient boosting tree classification, used for auxiliary determination of palatability level.

[0028] The fused feature vectors corresponding to the low-salt food sample to be evaluated and the reference sample are respectively input into the palatability quantification evaluation model to obtain the comprehensive palatability score of the low-salt food sample to be evaluated. P 1 Combined palatability score with reference sample P 0 The overall palatability score is converted to a range of 0 to 100 after standardization or normalization. A higher score indicates that the model predicts a higher overall palatability of the sample.

[0029] The palatability quantification evaluation model can also output at least one flavor dimension sub-score for the low-salt food sample to be evaluated, wherein the flavor dimension sub-score includes: ① Salty taste neurological response score; ② Umami nerve response score; ③ Oral stimulation sensory nerve response score; ④ Aroma nerve response score; ⑤ Pleasure level neural response score.

[0030] For different categories of low-salt foods, one or more of the flavor dimension sub-scores are selected as model auxiliary outputs based on their sensory attributes. For samples that do not have obvious oral irritation, aroma characteristics, or other specific flavor attributes, sub-scores for the corresponding dimensions may not be output.

[0031] In one implementation, the fusion feature vector of the training samples can be used as input, and one or more of the subject's preference selection results, overall liking score and purchase intention can be used as training labels to construct a neural preference evaluation model; the fusion feature vector of the low-salt food sample to be evaluated is input into the neural preference evaluation model to obtain a neural preference index of 0 to 100.

[0032] The flavor dimension sub-scores are used to reflect the model's prediction results for different perception dimensions and to help identify relatively weak flavor dimensions; the palatability comprehensive score is the comprehensive prediction result obtained by the palatability quantitative evaluation model based on the fused feature vector, and is not a simple arithmetic mean of the flavor dimension sub-scores.

[0033] In one implementation, the importance of built-in features or permutation importance analysis can be used to obtain the contribution of different EEG neural response features, event-related potential features and cerebral blood oxygenation response features to the prediction results of palatability comprehensive score or flavor dimension sub-score. Combined with the flavor dimension sub-scores, the main factors affecting the palatability of the sample can be identified to help generate product optimization suggestions.

[0034] S6. Palatability grade and determination of the match between salt reduction rate and palatability Based on the palatability comprehensive score obtained in step S5P 1 The palatability grade of the low-salt food samples to be evaluated is determined. In one embodiment, the palatability grade is divided as follows: P 1 ≥85, is judged as excellent; 70≤ P 1 <85, judged as acceptable; 60≤ P 1 <70, classified as requiring optimization; P 1 <60, classified as not recommended.

[0035] Calculate the actual salt reduction rate of the low-salt food sample to be evaluated based on its salt or sodium content compared to the reference sample. Rs, Actual salt reduction rate Rs according to Calculate, where, C 0 For reference, the salt or sodium content of the sample. C 1 The salt or sodium content of the low-salt food sample to be evaluated. C 0 and C 1 The same testing indicators, units of measurement, and calculation standards were used. The palatability retention rate was... R p according to Calculate, where, P 0 The overall palatability score is used as a reference sample. P 1 The overall palatability score for the low-salt food samples to be evaluated, and P 0 >0.

[0036] Let the target salt reduction rate of the low-salt food sample to be evaluated be... R t According to the actual salt reduction rate R s With the target salt reduction rate R t Relationship and palatability score P 1 and palatability retention rate R p The matching degree between salt reduction rate and palatability is determined. The matching degree between salt reduction rate and palatability includes high matching, medium matching, and low matching.

[0037] In one implementation: When the actual salt reduction rate R s Greater than or equal to the target salt reduction rate R t Overall palatability score P 1 Not less than 70 and palatability retention rate R p A score of 90% or higher is considered a high match. When the actual salt reduction rate R s A score greater than 0 indicates a good palatability score. P 1 Not less than 60 and palatability retention rate R p If the match rate is not less than 80%, but does not simultaneously meet the high match criteria, it is judged as a medium match. When the actual salt reduction rate R s Palatability score less than or equal to 0 P 1 Below 60 or palatability retention rate R p If the match rate is below 80%, it is considered a low match.

[0038] A high match indicates that the low-salt food sample being evaluated has achieved the target salt reduction rate and maintained a high level of palatability; a medium match indicates that the sample has a certain salt reduction effect and palatability retention ability, but the target salt reduction rate or some palatability indicators have not yet reached the high match condition and still need further optimization; a low match indicates that the sample has not achieved effective salt reduction, or the overall palatability score or palatability retention rate after salt reduction is low, and it is not suitable to be directly used as a candidate low-salt formula.

[0039] S7. Consistency Analysis of Subjective Evaluation and Neurological Preferences The subjects' subjective overall preference ratings for the low-salt food samples and the neural preference index obtained in step S5 were converted into standardized scores of 0 to 100, and the consistency between the two was analyzed.

[0040] In one implementation, according to Calculate the consistency index between subjective evaluation and neural preference. CI ,in, S To standardize subjective overall preference ratings, N This is the neural preference index. The consistency index ranges from 0 to 100; the higher the value, the greater the consistency between subjective evaluation and neural preference.

[0041] S8. Product optimization suggestions generated Based on the overall palatability score, palatability grade, salt reduction rate and palatability matching degree, subjective evaluation and neural preference consistency index, and flavor dimension sub-scores of the low-salt food samples to be evaluated, the flavor dimensions with relatively low scores are identified, and corresponding product optimization suggestions are generated.

[0042] When the salty taste neural response score is lower than the preset threshold, it indicates that the sample's salty taste perception or salty taste coordination is insufficient. Under the premise of meeting the target salt reduction requirements, the distribution of salt substitutes, salty taste enhancers, salty taste enhancing peptides, or table salt in the food matrix can be adjusted.

[0043] When the umami nerve response score is lower than the preset threshold, it indicates that the umami intensity or umami coordination of the sample is insufficient. The types and proportions of amino acid umami substances, nucleotide umami substances, yeast extracts or fermentation flavor components can be adjusted.

[0044] When the aroma nerve response score is lower than the preset threshold, it indicates that the aroma perception or aroma pleasure of the sample is insufficient. The aroma components, spice ratio, aroma release method or related processing conditions can be adjusted.

[0045] When the oral stimulation sensory nerve response score is higher than the preset appropriate range, it indicates that the spiciness, numbing sensation, burning sensation, cooling sensation or other oral stimulation of the sample is too strong. The proportion of the corresponding irritating ingredients can be reduced or the ratio between different irritating ingredients can be adjusted.

[0046] When the pleasure response score is below the preset threshold or the neural preference index is low, the flavor dimension sub-scores such as saltiness, umami, aroma and oral irritation can be combined to comprehensively optimize one or more of the following: salt content, flavor enhancement components, aroma components, irritating ingredients and processing conditions.

[0047] When the overall palatability score of the low-salt food sample to be evaluated reaches the acceptable level or above, and the salt reduction rate and palatability match well, the sample can be identified as a candidate low-salt formula; when the palatability level is to be optimized or the salt reduction rate and palatability match well or poorly, targeted optimization suggestions are generated based on the flavor dimension with the lower score.

[0048] The product optimization suggestions are used to assist in the screening and adjustment of low-salt food formulas, and do not directly limit the specific amount of each ingredient to be added; the types of ingredients, the addition ratios, and the processing conditions are determined according to the food category, the target salt reduction rate, and the product quality requirements.

[0049] This invention also provides a quantitative evaluation system for the palatability of low-salt foods based on multimodal brain neural signals, comprising: The sample management module is used to input basic information about the low-salt food samples to be evaluated and their corresponding reference samples. The basic information includes one or more of the following: food category, salt content, sodium content, formula information, processing method, storage conditions, and target consumer group. The module also inputs the target salt reduction rate of the low-salt food samples to be evaluated. The stimulus presentation module is used to control the presentation order, sample temperature, single intake, stimulus duration, stimulus interval, number of repetitions, and rinsing or taste recovery time of the low-salt food sample to be evaluated and the reference sample according to the preset sensory stimulus process, generate event markers corresponding to the sample presentation process, and collect the subject's subjective overall preference rating of the low-salt food sample to be evaluated. The signal acquisition module is used to simultaneously acquire the subject's electroencephalogram (EEG) signals and functional near-infrared cerebral oxygenation response signals during the presentation of sample stimuli, and record event markers; based on the event markers, the stimulation events are synchronized with the time of different modalities of brain nerve signals; The signal preprocessing module is used to preprocess the EEG signals and functional near-infrared cerebral oxygenation response signals, including filtering, artifact correction, baseline correction and abnormal data processing, and to obtain event-related potential responses from continuous EEG signals based on event markers. The feature extraction module is used to extract EEG neural response features, event-related potential features, and cerebral blood oxygenation response features related to the palatability of low-salt foods from preprocessed EEG signals, event-related potential responses, and functional near-infrared cerebral blood oxygenation response signals. The multimodal feature fusion module is used to process missing values, standardize, filter features, and splice features of the EEG neural response features, event-related potential features, and cerebral blood oxygenation response features to obtain the fused feature vectors corresponding to the low-salt food sample to be evaluated and the reference sample, respectively. The model analysis module is used to input the fusion feature vectors corresponding to the low-salt food sample to be evaluated and the reference sample into a pre-built palatability quantification evaluation model to obtain the comprehensive palatability score of the low-salt food sample to be evaluated and the reference sample, and to obtain the flavor dimension sub-score of the low-salt food sample to be evaluated; it is also used to input the fusion feature vector of the low-salt food sample to be evaluated into a pre-built neural preference evaluation model to obtain the neural preference index. The judgment output module is used to calculate the actual salt reduction rate based on the salt or sodium content of the low-salt food sample to be evaluated and the reference sample, calculate the palatability retention rate based on the overall palatability score of the low-salt food sample to be evaluated and the reference sample, calculate the consistency index between subjective evaluation and neural preference based on the overall subjective preference score and the neural preference index, and output one or more of the following: overall palatability score, palatability grade, flavor dimension sub-score, palatability retention rate, salt reduction rate and palatability matching degree, consistency index between subjective evaluation and neural preference, and product optimization suggestions.

[0050] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention reduces the reliance on a single subjective sensory score in the palatability evaluation of low-salt foods by synchronously collecting and fusing electroencephalogram (EEG) signals and functional near-infrared cerebral oxygenation response signals, thereby improving the objectivity, repeatability, and quantifiability of the evaluation results.

[0051] 2. This invention is not limited to a single saltiness intensity evaluation. It can integrate neural response features related to one or more of the sensory dimensions of saltiness, umami, oral stimulation, aroma, pleasure and overall acceptability, and output the corresponding flavor dimension sub-scores and palatability comprehensive scores.

[0052] 3. This invention can perform consistency analysis on the subject's overall subjective preference score and neural preference index, which helps to identify differences between subjective evaluation and neural response that are not easily found in traditional questionnaire surveys or artificial sensory evaluations.

[0053] 4. This invention can output a comprehensive palatability score, flavor dimension sub-score, palatability grade, palatability retention rate, and salt reduction rate and palatability matching degree, providing a quantitative basis for formula screening, flavor optimization and quality control of low-salt foods.

[0054] 5. This invention has a wide range of applications and can be used for palatability evaluation of various low-salt foods, such as low-salt seasonings, low-salt meat products, low-salt salt water products, low-salt convenience foods, low-salt snack foods, low-salt pre-cooked dishes, low-salt fermented foods, and low-salt compound seasoning foods.

[0055] 6. This invention has a standardized process for sample stimulus presentation, brain nerve signal acquisition, signal preprocessing, feature extraction, multimodal feature fusion, model analysis, and result judgment. It can be combined with EEG acquisition equipment, functional near-infrared detection equipment, food sensory evaluation system, and enterprise quality control platform, and has the potential to be applied to the research and development and quality control of low-salt foods. Attached Figure Description

[0056] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the embodiments are briefly described below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the described drawings without any creative effort.

[0057] Figure 1 This is a flowchart of the method for quantitatively evaluating the palatability of low-salt foods based on multimodal brain nerve signals, as described in this invention. Figure 2 This is a schematic diagram illustrating the standardized sensory stimulation presentation of low-salt foods and the synchronous acquisition of electroencephalogram (EEG) signals and functional near-infrared cerebral oxygenation response signals according to the present invention. Figure 3 This is a schematic diagram illustrating the multimodal brain nerve signal feature extraction, multimodal feature fusion, and palatability model construction of the present invention. Figure 4 This is a flowchart for determining the palatability grade of low-salt food and the matching degree between salt reduction rate and palatability in this invention. Figure 5 This invention presents a schematic diagram of the structure of a quantitative evaluation system for the palatability of low-salt foods based on multimodal brain neural signals. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The following embodiments are for illustrative purposes only and are not intended to limit the invention.

[0059] Example 1: Palatability Evaluation of Low-Salt Compound Seasoning 1. Sample setup Using a basic formula of a compound seasoning as the evaluation object, standard salt content control samples and three groups of low-salt treated samples were prepared according to the same ingredient composition and processing technology, with salt reduction rates of 10%, 20%, and 30%, respectively. Except for the amount of salt added, the other ingredient composition, addition ratio, and processing conditions were consistent across all groups. Sample grouping information is shown in Table 1.

[0060] Table 1. Grouping information of compound seasoning samples with different salt reduction gradients Sample group Relative amount of added salt Salt reduction rate Sample Description CK 100% 0% Standard salt content control group LS-10 90% 10% Mild salt reduction group LS-20 80% 20% Moderate salt reduction group LS-30 70% 30% Highly Salt-Reduced Group In this embodiment, 40 subjects were recruited. An event-related taste stimulation paradigm was used to simultaneously collect continuous EEG signals and functional near-infrared cerebral oxygenation response signals. Event-related potential features were extracted from the continuous EEG signals based on event markers.

[0061] The evaluation indicators include subjective sensory evaluation indicators and model output indicators. Subjective sensory evaluation indicators include saltiness intensity, umami harmony, aroma acceptability, and overall preference; model output indicators include saltiness neural response score, umami neural response score, aroma neural response score, pleasure neural response score, and palatability comprehensive score.

[0062] 2. Sensory evaluation and neural response scoring results Both subjective sensory evaluation indicators and model output indicators were converted to a score of 0–100. A higher subjective saltiness intensity score indicates a stronger perceived saltiness by the subject; higher scores for umami harmony, aroma acceptability, and overall liking indicate a higher degree of harmony or acceptability in the corresponding sensory dimensions. A higher model output score indicates a higher degree of acceptability or overall palatability in the corresponding flavor dimension predicted by the model. Results for each group are expressed as mean ± standard deviation, as shown in Table 2.

[0063] Table 2 Subjective evaluation and neural response quantitative indicators of compound seasonings with different salt reduction gradients sample Subjective saltiness rating Subjective overall preference Salty taste neurological response score Umami Nerve Response Score Aroma Nerve Response Score Pleasure Neurological Response Score Overall palatability score Palatability rating CK 88.6 ±4.2 86.3 ±5.1 90.2 ± 3.8 84.5 ± 4.7 82.6 ± 5.2 85.1 ± 4.6 86.4 ±3.9 Excellent LS-10 84.1 ±4.8 84.7 ±4.3 85.6 ± 4.1 85.2 ± 4.4 83.1 ± 4.9 84.8 ± 4.3 84.9 ±4.1 Acceptable level LS-20 78.5 ±5.3 81.2 ±5.6 78.9 ± 4.9 83.7 ± 4.6 82.4 ± 5.1 81.5 ± 4.8 81.7 ±4.5 Acceptable level LS-30 65.3 ±6.7 68.4 ±7.2 63.8 ± 6.1 76.2 ± 5.5 79.5 ± 5.8 67.1 ± 6.4 69.2 ±5.9 Optimization required 3. Results of the matching between salt reduction rate and palatability Using the overall palatability score of the control group CK as a reference, the palatability retention rate of each low-salt treatment sample was calculated. The matching degree between the salt reduction rate and palatability was determined by combining the actual salt reduction rate, the overall palatability score and the palatability retention rate. The results are shown in Table 3.

[0064] Table 3. Analysis Results of the Matching Degree between Salt Reduction Rate and Palatability of Low-Salt Compound Seasonings sample Salt reduction rate Overall palatability score Palatability retention rate Matching degree determination Product Development Recommendations LS-10 10% 84.9 98.3% High matching It can be used as a candidate solution for low-salt optimization. LS-20 20% 81.7 94.6% High matching Recommended as the preferred candidate solution LS-30 30% 69.2 80.1% Matching The perception and enjoyment of saltiness need to be improved. The results showed that the LS-20 sample with an actual salt reduction rate of 20% achieved a comprehensive palatability score of 81.7 points, with a palatability retention rate of 94.6%, maintaining good overall palatability while achieving salt reduction, and was therefore classified as a high match. The LS-30 sample with an actual salt reduction rate of 30% still had an aroma neural response score of 79.5 points, but its saltiness neural response score and pleasure neural response score decreased to 63.8 points and 67.1 points, respectively, suggesting that its reduced overall palatability may be mainly related to a decrease in saltiness perception and pleasure. The results of this embodiment demonstrate that the present invention can classify the palatability of low-salt foods and help identify the main flavor dimensions affecting product palatability.

[0065] 4. Model Performance Evaluation A fused feature vector, consisting of EEG neural response features, event-related potential features extracted from EEG signals, and functional near-infrared cerebral oxygenation response features, was used as the model input. Subjective overall preference ratings were used as the regression training labels to construct partial least squares regression, support vector regression, and random forest regression models, respectively. All models underwent training and evaluation using the same data partitioning and validation methods. Based on the model-predicted palatability scores, a level mapping was performed according to a preset palatability level threshold, and the level determination accuracy was calculated. Model performance is shown in Table 4.

[0066] Table 4 Performance of Multimodal Feature Palatability Score Prediction and Grade Determination Model Input features <![CDATA[Training set R 2 > <![CDATA[Test set R 2 > RMSE MAE Classification accuracy PLSR EEG characteristics + ERP characteristics + fNIRS characteristics 0.82 0.76 5.84 4.62 80.00% SVR EEG characteristics + ERP characteristics + fNIRS characteristics 0.88 0.83 4.71 3.86 85.00% Random Forest EEG characteristics + ERP characteristics + fNIRS characteristics 0.91 0.87 3.95 3.21 90.00% Note: ERP features are event-related potential features extracted from continuous EEG signals based on event labels.

[0067] The results show that the random forest regression model based on multimodal fusion features has good performance in predicting palatability scores, with an R² of 0.87, RMSE of 3.95, and MAE of 3.21 on the test set. After mapping the predicted scores to palatability levels, the accuracy of level determination is 90.0%. These results indicate that multimodal neural response features can be used for predicting palatability scores and determining levels for low-salt compound seasonings.

[0068] Example 2: Palatability evaluation of low-salt meat products 1. Sample setup Using a basic meat product formula as the evaluation object, standard salt content control samples and three low-salt treatment samples were prepared according to the same raw material composition and processing technology, with salt reduction rates of 15%, 25%, and 35%, respectively. Except for the amount of salt added, the raw material composition, addition ratio, and processing conditions were kept consistent across all groups. After processing, the samples were cooled to the preset consumption temperature and presented to the subjects with the same quality. Sample grouping information is shown in Table 5.

[0069] Table 5. Grouping Information for Low-Salt Meat Products Test Samples Sample group Relative amount of added salt Salt reduction rate Sample Description CK 100% 0% Standard salt content meat products M-LS15 85% 15% Lightly salted meat products M-LS25 75% 25% Moderately reduced salt meat products M-LS35 65% 35% Highly Salt-Reduced Meat Products This embodiment recruited 48 subjects. Samples for each group were presented according to a standardized sensory stimulation procedure. Continuous electroencephalogram (EEG) signals and functional near-infrared cerebral oxygenation response (FIR) signals were simultaneously collected. Event-related potential (ERP) features were extracted from the continuous EEG signals based on event markers. Evaluation indicators included subjective overall preference, repurchase intention, neuropreference index, and palatability score.

[0070] In this embodiment, the P300 peak amplitude serves as an event-related potential characteristic reflecting the food stimulus recognition and evaluation process; the prefrontal oxyhemoglobin response serves as a characteristic reflecting changes in prefrontal blood oxygen dynamics after sample stimulation; and the EEG β-band relative power serves as a characteristic reflecting EEG activity related to taste stimulus cognitive processing. These characteristics, along with other EEG and cerebral blood oxygen response characteristics, are used in conjunction with the model construction and are not used alone as direct criteria for determining flavor acceptance or consumer preference.

[0071] 2. Subjective evaluation and multimodal neural response data Subjective ratings, neuropreference index, and overall palatability rating were all converted to a score of 0–100. The results in the table are expressed as mean ± standard deviation. The overall palatability rating is a model output and not a simple average of the overall subjective liking and neuropreference index.

[0072] Table 6 Subjective evaluations and multimodal neural response indices of meat products with different salt reduction gradients sample Subjective overall preference Repurchase intention score P300 Peak Amplitude / μV Prefrontal HbO response index relative power in the β band Neuropreference Index Overall palatability score Palatability rating CK 87.4 ± 4.6 84.9 ±5.3 8.42 ± 1.16 0.91 ± 0.14 0.38 ±0.06 86.8 ±4.2 87.1 ± 4.0 Excellent M-LS15 85.6 ± 4.9 83.7 ±5.1 8.18 ± 1.21 0.88 ± 0.13 0.37 ±0.05 85.2 ±4.5 85.4 ± 4.2 Excellent M-LS25 79.8 ± 5.7 77.1 ±6.2 7.36 ± 1.25 0.79 ± 0.15 0.34 ±0.06 80.5 ±5.1 80.2 ± 4.8 Acceptable level M-LS35 70.2 ± 7.4 61.8 ±8.1 5.91 ± 1.42 0.61 ± 0.18 0.28 ±0.07 64.6 ±6.8 66.3 ± 6.2 Optimization required Note: The prefrontal HbO normalized response index and the relative power of the β band are both dimensionless indicators.

[0073] 3. Consistency Analysis of Subjective Evaluation and Neurological Preferences The subjective overall preference score and the neural preference index were converted into standardized scores ranging from 0 to 100. Let the subjective overall preference score be... S The neuropreference index is N absolute difference value D according to Calculate the consistency index CI according to The calculation shows that the smaller the absolute difference value and the higher the consistency index, the higher the degree of consistency between subjective evaluation and neural preference. In this embodiment, when... D When ≤3, it is judged as high consistency; when 3 < D When ≤5, it is judged as consistent; when D A score >5 indicates low consistency. The results of the consistency analysis between subjective evaluations and neural preferences for low-salt meat products are shown in Table 7.

[0074] Table 7. Results of Consistency Analysis of Subjective Evaluation and Neurological Preferences for Low-Salt Meat Products sample Subjective overall preference Neuropreference Index absolute difference value Consistency Index Consistency determination CK 87.4 86.8 0.6 99.4 High consistency M-LS15 85.6 85.2 0.4 99.6 High consistency M-LS25 79.8 80.5 0.7 99.3 High consistency M-LS35 70.2 64.6 5.6 94.4 Low consistency For the M-LS35 sample with a 35% reduction in salt, the overall subjective preference score was 70.2 points, and the neural preference index was 64.6 points, with an absolute difference of 5.6, indicating low consistency. This sample also showed lower P300 peak amplitude, prefrontal HbO response, and β-band relative power compared to the control group. These results suggest that consistency analysis between subjective evaluation and neural preference can supplement the reflection of neural response differences not fully captured by a single subjective sensory score.

[0075] 4. Performance comparison of models with different feature combinations A palatability rating prediction model was constructed using basic EEG features, basic EEG features combined with event-related potential (ERP) features, basic EEG features combined with functional near-infrared cerebral oxygenation response (FIR) features, and all multimodal fusion features as model inputs. Based on the comprehensive palatability score predicted by the model, a level mapping was performed according to a preset palatability level threshold, and the level determination accuracy was calculated. EEP features were extracted from continuous EEG signals and were not considered as an independent acquisition modality.

[0076] Table 8. Performance comparison of palatability evaluation models under different feature combinations Model Input feature combination Test set R² RMSE MAE Accuracy of rating determination Basic EEG Feature Model EEG basic characteristics 0.74 6.28 5.03 77.10% EEG and ERP Feature Model EEG basic characteristics + ERP characteristics 0.81 5.36 4.31 83.30% EEG and fNIRS fusion feature model EEG basic features + fNIRS features 0.85 4.72 3.84 87.50% Multimodal fusion feature model EEG basic features + ERP features + fNIRS features 0.90 3.89 3.12 91.70% Compared with models using only basic EEG features, the addition of event-related potential (ERP) features and functional near-infrared cerebral oxygenation response (FIR) features improved both the predictive and grading performance of palatability scores. The model using all fused features achieved an R² of 0.90, an RMSE of 3.89, a MAE of 3.12, and a grading accuracy of 91.7% on the test set. These results indicate that fusing basic EEG features, EEP features extracted from EEG signals, and FIR features helps improve the predictive performance of palatability evaluation models for low-salt meat products.

[0077] Example 3: Application of formula optimization in the research and development of low-salt foods 1. Experimental Objective Using the basic formula of pre-made seasoning sauce as the evaluation object, the palatability of samples under different salt reduction ratios and flavor compensation strategies was evaluated using the method of this invention, and candidate formulas with both high actual salt reduction rate and high overall palatability were screened.

[0078] 2. Formulation Design Using the standard salt content formula as a control group, five low-salt formulas were prepared under the same basic formula and processing conditions. Each formula was adjusted according to the amount of salt added and flavor compensation components, while other raw material composition and processing conditions remained consistent. The salt reduction ratio and flavor compensation strategy for each group are shown in Table 9.

[0079] Table 9. Experimental Formulation Design Information for Low-Salt Pre-Cooked Vegetable Seasoning Sauces Sample number Actual salt reduction rate Flavor compensation strategy Formula instructions CK 0% none Standard salt content control formula <![CDATA[F1]]> 20% No compensation Only reduce the amount of added salt <![CDATA[F2]]> 20% Umami compensation Add yeast extract and flavor nucleotides <![CDATA[F3]]> 25% Umami and aroma compensation Add yeast extract, flavor nucleotides and meat flavor base <![CDATA[F4]]> 30% Umami and spice compensation Added yeast extract, flavor nucleotides, pepper and Sichuan pepper extracts <![CDATA[F5]]> 35% Composite compensation Added yeast extract, flavor nucleotides, meat flavor base and salt-enhancing peptides In this embodiment, 40 subjects were recruited, and samples were presented to each group according to a randomized, blinded, and standardized sensory stimulation procedure. The presentation temperature, single intake, stimulation duration, stimulation interval, number of repetitions, and rinsing time were kept consistent for each sample.

[0080] 3. Subjective sensory rating results Subjective sensory evaluation indicators include saltiness balance, umami intensity, aroma acceptability, oral stimulation balance, and overall preference. Each indicator is evaluated using a score from 0 to 100. A higher umami intensity score indicates a stronger perceived umami flavor by the subject; higher scores for saltiness balance, aroma acceptability, oral stimulation balance, and overall preference indicate a higher degree of balance or acceptability in the corresponding sensory dimensions.

[0081] Table 10 Subjective sensory evaluation results of different low-salt seasoning sauces Sample number Saltiness balance Umami intensity Aroma acceptance Oral stimulation coordination Overall Preference CK 88.2 ± 4.1 83.5 ± 4.8 84.6 ± 5.2 82.4 ± 4.9 86.1 ± 4.5 <![CDATA[F1]]> 70.4 ± 6.8 74.1 ± 6.2 80.3 ± 5.6 81.5 ± 5.1 72.8 ± 6.4 <![CDATA[F2]]> 79.6 ± 5.3 84.2 ± 4.7 81.7 ± 5.3 80.9 ± 5.4 81.5 ± 5.1 <![CDATA[F3]]> 82.7 ± 4.9 85.6 ± 4.4 86.8 ± 4.6 81.6 ± 5.0 85.2 ± 4.3 <![CDATA[F4]]> 76.8 ± 5.7 84.9 ± 4.8 84.1 ± 5.0 74.6 ± 6.1 78.4 ± 5.8 <![CDATA[F5]]> 73.5 ± 6.2 83.8 ± 5.1 85.4 ± 4.9 72.8 ± 6.5 76.9 ± 6.0 4. Multimodal brain neural response characteristics and model output During the presentation of the sample stimulus, continuous EEG signals and functional near-infrared cerebral oxygenation response signals were acquired simultaneously, and event-related potential features were extracted from the continuous EEG signals based on event markers.

[0082] Representative neural response indicators extracted included P300 peak amplitude, prefrontal HbO response, and EEG β-band relative power. These neural response indicators, along with other EEG neural response features, event-related potential features, and cerebral blood oxygenation response features, were used in model analysis to obtain aroma neural response scores and neural preference indices. Both aroma neural response scores and neural preference indices were normalized to 0–100 points.

[0083] Table 11 Representative neural response indices and model outputs for different low-salt seasoning sauces Sample number P300 Peak Amplitude / μV Prefrontal HbO response relative power in the β band Aroma Nerve Response Score Neuropreference Index CK 8.31 ± 1.12 0.89 ± 0.13 0.37 ± 0.05 84.9 ± 4.7 86.4 ± 4.1 <![CDATA[F1]]> 5.86 ± 1.34 0.58 ± 0.16 0.28 ± 0.07 77.2 ± 5.8 68.7 ± 6.5 <![CDATA[F2]]> 7.41 ± 1.19 0.76 ± 0.14 0.33 ± 0.06 80.6 ± 5.2 80.8 ± 5.0 <![CDATA[F3]]> 8.05 ± 1.08 0.85 ± 0.12 0.36 ± 0.05 86.2 ± 4.5 85.9 ± 4.2 <![CDATA[F4]]> 6.82 ± 1.27 0.71 ± 0.15 0.31 ± 0.06 83.1 ± 5.0 77.6 ± 5.7 <![CDATA[F5]]> 6.47 ± 1.31 0.66 ± 0.17 0.30 ± 0.07 84.0 ± 5.2 75.4 ± 6.1 Note: The prefrontal HbO response is a standardized response exponent, which is dimensionless; the relative power in the β band is a dimensionless indicator.

[0084] 5. Overall palatability score and formula screening results A pre-built and validated palatability quantification evaluation model was used to obtain a comprehensive palatability score based on the multimodal fusion feature vectors of each sample. This comprehensive palatability score is a model prediction result and not a simple arithmetic mean of subjective overall preference and neural preference index.

[0085] Using the overall palatability score of the CK group as a reference, the palatability retention rate of each low-salt formula was calculated, and the matching degree between the salt reduction rate and palatability was determined by combining the overall palatability score, the actual salt reduction rate, and the palatability retention rate.

[0086] Table 12. Overall Palatability Scores and Screening Results of Different Low-Salt Seasoning Sauces Sample number Actual salt reduction rate Overall palatability score Palatability rating Palatability retention rate Salt reduction rate and palatability matching Judgment result CK 0% 86.3 Excellent 100.0% not applicable control group <![CDATA[F1]]> 20% 70.5 Acceptable level 81.7% Matching Needs optimization <![CDATA[F2]]> 20% 81.2 Acceptable level 94.1% High matching Acceptable candidate formulations <![CDATA[F3]]> 25% 85.6 Excellent 99.2% High matching Preferred candidate formulations <![CDATA[F4]]> 30% 78.0 Acceptable level 90.4% High matching Oral stimulation still needs optimization <![CDATA[F5]]> 35% 76.1 Acceptable level 88.2% Matching Not the first choice 6. Palatability shortcomings analysis of each formulation Table 13 Palatability shortcomings and optimization suggestions for different low-sodium sauce formulations Sample number Main advantages Main shortcomings Optimization suggestions <![CDATA[F1]]> Salt addition reduced by 20% Saltiness has low balance and low neurological preference index. Increase umami-compensating components or optimize saltiness-enhancing components <![CDATA[F2]]> It has good umami intensity and overall palatability. Aroma acceptance and aroma neurological response scores were still lower than CK. Optimize meat aroma or complex aroma components <![CDATA[F3]]> The saltiness, umami, aroma, nerve preferences, and overall palatability are all close to CK. No obvious shortcomings in palatability were found. As a preferred candidate formulation for subsequent scale-up verification <![CDATA[F4]]> The salt content is significantly reduced, thus preserving the umami flavor well. Low coordination of oral stimulation Reduce the proportion of pepper or Sichuan pepper extract and optimize their ratio. <![CDATA[F5]]> The salt content was reduced the most, and the aroma was more acceptable. The balance between saltiness and oral stimulation is low. Appropriately reduce the amount of salt or optimize the ratio of salt-enhancing peptides and irritating ingredients. 7. Conclusion of the Example The results showed that the overall palatability score of the control formula (CK) was 86.3. Formula F1, which only reduced salt without flavor compensation, saw its overall palatability score drop to 70.5, with a palatability retention rate of 81.7%. Its saltiness harmony and neurobiological preference index were both lower than the control formula, indicating that simply reducing salt content may affect the overall palatability of the product. Formula F2, which incorporated umami compensation, saw its overall palatability score increase to 81.2, with a palatability retention rate of 94.1%, demonstrating that umami compensation helps improve the flavor deficiencies caused by salt reduction.

[0087] Among the various low-salt formulations, formulation F3 achieved a comprehensive palatability score of 85.6, a palatability retention rate of 99.2%, and a neuropreference index of 85.9, all of which are close to the standard salt content control formulation. The results show that combined compensation using umami components and meat flavor base helps to better maintain the overall palatability of the product. Therefore, F3 was identified as the preferred candidate formulation, and its salt reduction rate and palatability were judged to be a high match.

[0088] Formula F4 achieved an actual salt reduction rate of 30%, with a comprehensive palatability score of 78.0 and a palatability retention rate of 90.4%, classifying it as a high match according to preset rules. However, its oral irritation harmony was relatively low, suggesting that the proportion of pepper and Sichuan pepper extracts added still needs further optimization. Formula F5 achieved an actual salt reduction rate of 35%, with a comprehensive palatability score of 76.1. Both its saltiness harmony and oral irritation harmony were lower than those of Formula F3, indicating that even with a complex flavor compensation strategy, palatability may still decline at a higher actual salt reduction rate.

[0089] The above results show that the method of the present invention can quantitatively evaluate the palatability of different low-salt food formulations, identify the main flavor dimensions affecting palatability, output palatability grades and the matching degree between salt reduction rate and palatability, and provide quantitative basis for the screening and optimization of low-salt food formulations.

Claims

1. A method for quantitatively evaluating the palatability of low-salt foods based on multimodal brain neural signals, characterized in that, Includes the following steps: S1. Obtain the low-salt food sample to be evaluated and its corresponding reference sample, record the salt or sodium content of the low-salt food sample to be evaluated and the reference sample, as well as the target salt reduction rate of the low-salt food sample to be evaluated, and present the low-salt food sample to be evaluated and the reference sample to the subject according to the preset sensory stimulation process. S2. During the sensory stimulation process, the subject's electroencephalogram (EEG) signal and functional near-infrared cerebral oxygenation response signal are collected simultaneously, and at least one time point in the sample presentation, smelling, ingestion, chewing, swallowing and evaluation is recorded by event tagging. S3. Preprocess the EEG signal and functional near-infrared cerebral oxygenation response signal, and extract the EEG neural response features, event-related potential features obtained from the EEG signal, and cerebral oxygenation response features. S4. Standardize, select features, and fuse features at the level of brain electrical neural response features, event-related potential features, and brain blood oxygenation response features to obtain the fused feature vectors corresponding to the low-salt food sample to be evaluated and the reference sample. S5. Input the fusion feature vectors corresponding to the low-salt food sample to be evaluated and the reference sample into the pre-constructed palatability quantitative evaluation model to obtain the comprehensive palatability score of the low-salt food sample to be evaluated and the comprehensive palatability score of the reference sample, and obtain at least one flavor dimension sub-score of the low-salt food sample to be evaluated. S6. Calculate the actual salt reduction rate based on the salt content or sodium content of the low-salt food sample to be evaluated and the reference sample, calculate the palatability retention rate based on the comprehensive palatability score of the low-salt food sample to be evaluated and the reference sample, and output the palatability evaluation result based on the comprehensive palatability score, the actual salt reduction rate and the palatability retention rate. The palatability evaluation results include at least a comprehensive palatability score and a match between salt reduction rate and palatability, and include at least one of palatability grade and flavor dimension sub-scores.

2. The method according to claim 1, characterized in that, The low-salt food sample to be evaluated in step S1 is selected from any one of low-salt condiments, low-salt meat products, low-salt salt products, low-salt convenience foods, low-salt snack foods, low-salt pre-cooked dishes, low-salt fermented foods, and low-salt compound seasoning foods; the reference sample is a standard salt content food sample, a conventional formula food sample, a commercially available food sample of the same category, or a benchmark sample selected from food samples with a preset salt reduction rate gradient that has the same or similar food matrix as the low-salt food sample to be evaluated.

3. The method according to claim 1, characterized in that, The preset sensory stimulation process described in step S1 includes one or more of the following: sample coding, randomization, blind presentation, sample temperature control, sample intake control, stimulation duration control, stimulation interval control, mouth rinsing or taste recovery control, and blank baseline acquisition.

4. The method according to claim 1, characterized in that, The event-related potential characteristics mentioned in step S3 are obtained by performing event segmentation, baseline correction, abnormal trial elimination, and superposition averaging on the EEG signals labeled with the event.

5. The method according to claim 1, characterized in that, The EEG neural response features mentioned in step S3 include one or more of the following: EEG power spectrum features, time-frequency features, EEG coherence features, functional connectivity features, frontal lobe asymmetry features, and EEG complexity features; the event-related potential features include one or more of the following: event-related potential peak amplitude, peak latency, and average amplitude within a preset time window; the brain blood oxygenation response features include one or more of the following: changes in oxyhemoglobin concentration, changes in deoxyhemoglobin concentration, blood oxygenation response peak, response area, response slope, and brain region connectivity features.

6. The method according to claim 1, characterized in that, The feature-level fusion in step S4 includes processing missing values, standardizing, selecting features, and concatenating features of the EEG neural response features, event-related potential features, and cerebral blood oxygenation response features to obtain a fused feature vector.

7. The method according to claim 1, characterized in that, The palatability quantification evaluation model described in step S5 uses the fusion feature vector of the training samples as input and constructs it using one or more of the following as training labels: subjective sensory rating, liking rating, preference selection result, and purchase intention. The palatability quantification evaluation model is a regression model or a combination model consisting of a regression model and a classification model, and is selected from one or more of the following: partial least squares regression model, support vector machine model, random forest model, and gradient boosting tree model.

8. The method according to claim 1, characterized in that, The actual salt reduction rate mentioned in step S6 Rs according to Calculate, where, C 0 For reference, the salt or sodium content of the sample. C 1 The salt or sodium content of the low-salt food sample to be evaluated, and C 0 and C 1 Using the same testing indicators and units of measurement; the palatability retention rate R p according to Calculate, where, P 0 The overall palatability score is used as a reference sample. P 1 The overall palatability score is used to evaluate the low-salt food samples. The matching degree between the salt reduction rate and palatability is determined based on the actual salt reduction rate, the target salt reduction rate, the overall palatability score, and the palatability retention rate.

9. The method according to claim 1, characterized in that, The method further includes: collecting the subjects' subjective overall preference scores for the low-salt food samples to be evaluated; inputting the fusion feature vector of the low-salt food samples to be evaluated into a pre-constructed neural preference evaluation model to obtain a neural preference index; converting the subjective overall preference scores and neural preference index into standardized scores of 0-100 points respectively, and then... Calculate the consistency index between subjective evaluation and neural preference. CI ,in, S To standardize subjective overall preference ratings, N The neural preference index is used; based on the consistency index and flavor dimension sub-score, optimization suggestions for low-salt food formulations are output.

10. A quantitative evaluation system for the palatability of low-salt foods based on multimodal brain neural signals, characterized in that, include: The sample management module is used to input the low-salt food sample to be evaluated, the reference sample, the salt or sodium content of the low-salt food sample to be evaluated and the reference sample, and the target salt reduction rate of the low-salt food sample to be evaluated. The stimulus presentation module is used to control the presentation order, stimulus duration, stimulus interval and repetition number of the low-salt food sample to be evaluated and the reference sample, and to collect the subject's subjective overall preference rating of the low-salt food sample to be evaluated. The signal acquisition module is used to simultaneously acquire the subject's electroencephalogram (EEG) signals and functional near-infrared cerebral oxygenation response signals, and record event markers; The signal preprocessing module is used to preprocess the electroencephalogram (EEG) signals and the functional near-infrared cerebral oxygenation response signals. The feature extraction module is used to extract brain electrical neural response features, event-related potential features, and brain blood oxygenation response features; The multimodal feature fusion module is used to process missing values, standardize, filter features, and splice features of the EEG neural response features, event-related potential features, and cerebral blood oxygenation response features to obtain the fused feature vectors corresponding to the low-salt food sample to be evaluated and the reference sample, respectively. The model analysis module is used to input the fusion feature vectors corresponding to the low-salt food sample to be evaluated and the reference sample into the palatability quantification evaluation model to obtain the comprehensive palatability score of the low-salt food sample to be evaluated and the reference sample, and to obtain the flavor dimension sub-score of the low-salt food sample to be evaluated; it is also used to input the fusion feature vector of the low-salt food sample to be evaluated into the neural preference evaluation model to obtain the neural preference index. The judgment output module is used to calculate the actual salt reduction rate based on the salt or sodium content of the low-salt food sample to be evaluated and the reference sample, calculate the palatability retention rate based on the overall palatability score of the low-salt food sample to be evaluated and the reference sample, calculate the consistency index between subjective evaluation and neural preference based on the overall subjective preference score and the neural preference index, and output the overall palatability score, palatability grade, flavor dimension sub-score, palatability retention rate, salt reduction rate and palatability matching degree, subjective evaluation and neural preference consistency index, and product optimization suggestions.

Images