Claudia Gonzalez Viejo

Development of a robotic pourer constructed with ubiquitous materials, open hardware and sensors to assess beer foam quality using computer vision and pattern recognition algorithms: RoboBEER

There are currently no standardized objective measures to assess beer quality based on the most significant parameters related to the first impression from consumers, which are visual characteristics of foamability, beer color and bubble size. This study describes the development of an affordable and robust robotic beer pourer using low-cost sensors, Arduino® boards, Lego® building blocks and servo motors for prototyping. The RoboBEER is also coupled with video capture capabilities (iPhone 5S) and automated post hoc computer vision analysis algorithms to assess different parameters based on foamability, bubble size, alcohol content, temperature, carbon dioxide release and beer color. Results have shown that parameters obtained from different beers by only using the RoboBEER can be used for their classification according to quality and fermentation type. Results were compared to sensory analysis techniques using principal component analysis (PCA) and artificial neural networks (ANN) techniques. The PCA from RoboBEER data explained 73% of variability within the data. From sensory analysis, the PCA explained 67% of the variability and combining RoboBEER and Sensory data, the PCA explained only 59% of data variability. The ANN technique for pattern recognition allowed creating a classification model from the parameters obtained with RoboBEER, achieving 92.4% accuracy in the classification according to quality and fermentation type, which is consistent with the PCA results using data only from RoboBEER. The repeatability and objectivity of beer assessment offered by the RoboBEER could translate into the development of an important practical tool for food scientists, consumers and retail companies to determine differences within beers based on the specific parameters studied.

Integration of non-invasive biometrics with sensory analysis techniques to assess acceptability of beer by consumers

Traditional sensory tests rely on conscious and self-reported responses from participants. The integration of non-invasive biometric techniques, such as heart rate, body temperature, brainwaves and facial expressions can gather more information from consumers while tasting a product. The main objectives of this study were i) to assess significant differences between beers for all conscious and unconscious responses, ii) to find significant correlations among the different variables from the conscious and unconscious responses and iii) to develop a model to classify beers according to liking using only the unconscious responses. For this study, an integrated camera system with video and infrared thermal imagery (IRTI), coupled with a novel computer application was used. Videos and IRTI were automatically obtained while tasting nine beers to extract biometrics (heart rate, temperature and facial expressions) using computer vision analysis. Additionally, an EEG mobile headset was used to obtain brainwave signals during beer consumption. Consumers assessed foam, color, aroma, mouthfeel, taste, flavor and overall acceptability of beers using a 9-point hedonic scale with results showing a higher acceptability for beers with higher foamability and lower bitterness. i) There were non-significant differences among beers for the emotional and physiological responses, however, significant differences were found for the cognitive and self-reported responses. ii) Results from principal component analysis explained 65% of total data variability and, along with the covariance matrix (p < 0.05), showed that there are correlations between the sensory responses of participants and the biometric data obtained. There was a negative correlation between body temperature and liking of foam height and stability, and a positive correlation between theta signals and bitterness. iii) Artificial neural networks were used to develop three models with high accuracy to classify beers according to level of liking (low and high) of three sensory descriptors: carbonation mouthfeel (82%), flavor (82%) and overall liking (81%). The integration of both sensory and biometric responses for consumer acceptance tests showed to be a reliable tool to be applied to beer tasting to obtain more information from consumers physiology, behavior and cognitive responses.

Assessment of beer quality based on foamability and chemical composition using computer vision algorithms, near infrared spectroscopy and machine learning algorithms: Objective assessment of beer quality

BACKGROUND Beer quality is mainly defined by its colour, foamability and foam stability, which are influenced by the chemical composition of the product such as proteins, carbohydrates, pH and alcohol. Traditional methods to assess specific chemical compounds are usually time-consuming and costly. This study used rapid methods to evaluate 15 foam and colour-related parameters using a robotic pourer (RoboBEER) and chemical fingerprinting using near infrared spectroscopy (NIR) from six replicates of 21 beers from three types of fermentation. Results from NIR were used to create partial least squares regression (PLS) and artificial neural networks (ANN) models to predict four chemometrics such as pH, alcohol, Brix and maximum volume of foam. RESULTS The ANN method was able to create more accurate models (R2  = 0.95) compared to PLS. Principal components analysis using RoboBEER parameters and NIR overtones related to protein explained 67% of total data variability. Additionally, a sub-space discriminant model using the absorbance values from NIR wavelengths resulted in the successful classification of 85% of beers according to fermentation type. CONCLUSION The method proposed showed to be a rapid system based on NIR spectroscopy and RoboBEER outputs of foamability that can be used to infer the quality, production method and chemical parameters of beer with minimal laboratory equipment.

Assessment of Beer Quality Based on a Robotic Pourer, Computer Vision, and Machine Learning Algorithms Using Commercial Beers: Machine learning for beer quality…

Sensory attributes of beer are directly linked to perceived foam-related parameters and beer color. The aim of this study was to develop an objective predictive model using machine learning modeling to assess the intensity levels of sensory descriptors in beer using the physical measurements of color and foam-related parameters. A robotic pourer (RoboBEER), was used to obtain 15 color and foam-related parameters from 22 different commercial beer samples. A sensory session using quantitative descriptive analysis (QDA® ) with trained panelists was conducted to assess the intensity of 10 beer descriptors. Results showed that the principal component analysis explained 64% of data variability with correlations found between foam-related descriptors from sensory and RoboBEER such as the positive and significant correlation between carbon dioxide and carbonation mouthfeel (R = 0.62), correlation of viscosity to sensory, and maximum volume of foam and total lifetime of foam (R = 0.75, R = 0.77, respectively). Using the RoboBEER parameters as inputs, an artificial neural network (ANN) regression model showed high correlation (R = 0.91) to predict the intensity levels of 10 related sensory descriptors such as yeast, grains and hops aromas, hops flavor, bitter, sour and sweet tastes, viscosity, carbonation, and astringency. PRACTICAL APPLICATIONS This paper is a novel approach for food science using machine modeling techniques that could contribute significantly to rapid screenings of food and brewage products for the food industry and the implementation of Artificial Intelligence (AI). The use of RoboBEER to assess beer quality showed to be a reliable, objective, accurate, and less time-consuming method to predict sensory descriptors compared to trained sensory panels. Hence, this method could be useful as a rapid screening procedure to evaluate beer quality at the end of the production line for industry applications.

Non-Contact Heart Rate and Blood Pressure Estimations from Video Analysis and Machine Learning Modelling Applied to Food Sensory Responses: A Case Study for Chocolate

Traditional methods to assess heart rate (HR) and blood pressure (BP) are intrusive and can affect results in sensory analysis of food as participants are aware of the sensors. This paper aims to validate a non-contact method to measure HR using the photoplethysmography (PPG) technique and to develop models to predict the real HR and BP based on raw video analysis (RVA) with an example application in chocolate consumption using machine learning (ML). The RVA used a computer vision algorithm based on luminosity changes on the different RGB color channels using three face-regions (forehead and both cheeks). To validate the proposed method and ML models, a home oscillometric monitor and a finger sensor were used. Results showed high correlations with the G color channel (R2 = 0.83). Two ML models were developed using three face-regions: (i) Model 1 to predict HR and BP using the RVA outputs with R = 0.85 and (ii) Model 2 based on time-series prediction with HR, magnitude and luminosity from RVA inputs to HR values every second with R = 0.97. An application for the sensory analysis of chocolate showed significant correlations between changes in HR and BP with chocolate hardness and purchase intention.

Assessment of beer quality based on foamability and chemical composition using computer vision algorithms, near infrared spectroscopy and artificial neural networks modelling techniques

BACKGROUND Beer quality is mainly defined by its colour, foamability and foam stability, which are influenced by the chemical composition of the product such as proteins, carbohydrates, pH and alcohol. Traditional methods to assess specific chemical compounds are usually time-consuming and costly. This study used rapid methods to evaluate 15 foam and colour-related parameters using a robotic pourer (RoboBEER) and chemical fingerprinting using near infrared spectroscopy (NIR) from six replicates of 21 beers from three types of fermentation. Results from NIR were used to create partial least squares regression (PLS) and artificial neural networks (ANN) models to predict four chemometrics such as pH, alcohol, Brix and maximum volume of foam. RESULTS The ANN method was able to create more accurate models (R2  = 0.95) compared to PLS. Principal components analysis using RoboBEER parameters and NIR overtones related to protein explained 67% of total data variability. Additionally, a sub-space discriminant model using the absorbance values from NIR wavelengths resulted in the successful classification of 85% of beers according to fermentation type. CONCLUSION The method proposed showed to be a rapid system based on NIR spectroscopy and RoboBEER outputs of foamability that can be used to infer the quality, production method and chemical parameters of beer with minimal laboratory equipment.

Assessment of beer quality based on foamability and chemical composition using computer vision algorithms, near infrared spectroscopy and machine learning algorithms

BACKGROUND Beer quality is mainly defined by its colour, foamability and foam stability, which are influenced by the chemical composition of the product such as proteins, carbohydrates, pH and alcohol. Traditional methods to assess specific chemical compounds are usually time-consuming and costly. This study used rapid methods to evaluate 15 foam and colour-related parameters using a robotic pourer (RoboBEER) and chemical fingerprinting using near infrared spectroscopy (NIR) from six replicates of 21 beers from three types of fermentation. Results from NIR were used to create partial least squares regression (PLS) and artificial neural networks (ANN) models to predict four chemometrics such as pH, alcohol, Brix and maximum volume of foam. RESULTS The ANN method was able to create more accurate models (R2  = 0.95) compared to PLS. Principal components analysis using RoboBEER parameters and NIR overtones related to protein explained 67% of total data variability. Additionally, a sub-space discriminant model using the absorbance values from NIR wavelengths resulted in the successful classification of 85% of beers according to fermentation type. CONCLUSION The method proposed showed to be a rapid system based on NIR spectroscopy and RoboBEER outputs of foamability that can be used to infer the quality, production method and chemical parameters of beer with minimal laboratory equipment.

Development of a Biosensory Computer Application to Assess Physiological and Emotional Responses from Sensory Panelists

In sensory evaluation, there have been many attempts to obtain responses from the autonomic nervous system (ANS) by analyzing heart rate, body temperature, and facial expressions. However, the methods involved tend to be intrusive, which interfere with the consumers’ responses as they are more aware of the measurements. Furthermore, the existing methods to measure different ANS responses are not synchronized among them as they are measured independently. This paper discusses the development of an integrated camera system paired with an Android PC application to assess sensory evaluation and biometric responses simultaneously in the Cloud, such as heart rate, blood pressure, facial expressions, and skin-temperature changes using video and thermal images acquired by the integrated system and analyzed through computer vision algorithms written in Matlab®, and FaceReaderTM. All results can be analyzed through customized codes for multivariate data analysis, based on principal component analysis and cluster analysis. Data collected can be also used for machine-learning modeling based on biometrics as inputs and self-reported data as targets. Based on previous studies using this integrated camera and analysis system, it has shown to be a reliable, accurate, and convenient technique to complement the traditional sensory analysis of both food and nonfood products to obtain more information from consumers and/or trained panelists.

Cross-cultural effects of food product familiarity on sensory acceptability and non-invasive physiological responses of consumers

This research evaluated the effects of product familiarity on the sensory acceptability and physiological responses of consumers toward different food stimuli using two populations (Asian vs. Western). Two studies were conducted: (1) an online questionnaire and (2) a tasting session. For (1), n = 102 (60% Asians and 40% Westerners) evaluated 31 food items visually for familiarity and liking whereas for (2), participants (n = 60; 48% Asians and 52% Westerners) evaluated 10 different foods (tortoise jelly, chili slices, beef jerky, dried tofu, Vegemite®, durian cake, octopus chips, chocolate, corn chips, and wasabi coated peas) by tasting for familiarity and liking (visual/aroma/taste/texture/overall). A novel Android® app (Bio-sensory App) was used to capture sensory and non-invasive physiological responses (temperature, heart rate and facial expressions) of consumers. In (1), Asian and Western participants differed in their familiarity scores, visual liking ratings, and the selection of emotion terms for the stimuli. In (2), cultural differences affected familiarity and the liking scores of appearance, aroma, taste and texture of the products. While food stimuli marginally affected the physiological responses of consumers for both cultures, Asian participants elicited higher temperature values compared to those of Westerners. Both studies (1 and 2) showed that familiarity of food products was positively associated to sensory liking for both cultural groups. These findings are useful to understand consumers acceptability based on both sensory and physiological responses.

Development of emotion lexicons to describe chocolate using the Check-All-That-Apply (CATA) methodology across Asian and Western groups

Emotion-based terms selected by Asians and Westerners were analyzed to develop lexicons associated with chocolate consumption. Hence, an online-based questionnaire (Study 1: N = 206; 51% Asians, 49% Westerners) and a chocolate (milk and dark) tasting session (Study 2: N = 75; 52% Asians, 48% westerners) were conducted to assess emotion terms related to chocolate consumption using Check-All-That-Apply methodology. Emotional satisfaction was the main reason for chocolate consumption. Furthermore, selection of emotional terms was different between cultures and gender. For both studies (1 and 2), flavor of chocolate was the most important factor that determined purchase intention. For Study 2, milk and dark chocolate evoked different emotion terms for participants. The lexicon developed for milk chocolate had similar emotion terms compared to the Study 1 lexicon (online). Developing an emotion lexicon using an online survey could provide a reduced lexicon compared to lexicons generated during the tasting session and can be used as a fast-screening method to develop simplified emotion lexicons due to its similarity to the tasting lexicon. Newly developed lexicons from this study can be applied to sensory consumer tests of chocolate.