Aroma detection systems for food and beverage and conversion of detected aromas to natural language descriptors
Inventors
Ivanov, Ilia N. • Muckley, Eric S. • Reina, Reinaldo C.
Assignees
UT Battelle LLC • University of Tennessee Research Foundation
Publication Number
US-12005649-B2
Publication Date
2024-06-11
Expiration Date
2041-09-23
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Abstract
A system for determining an age and/or quality of food or beverage based on one or more combinations of outputs from gas sensors input into a deployed machine learning model is provided. The system may comprise an electronic nose which may comprise a housing and the gas sensors. The housing may have an air channel. Each sensor has its active sensor portion in the air channel. A system for predicting one or more natural language descriptors associated with aromas of an item based on one or more outputs of the gas sensors and calculated one or more ratios input into a logistic regression model is also provided.
Core Innovation
The invention discloses a system for predicting one or more analytes, age, or quality of food or beverage based on outputs from a plurality of thin film gas sensors integrated in an electronic nose (e-nose). The system utilizes machine learning models that are trained and tested using datasets derived from sensor outputs corresponding to various analytes or food/beverage conditions. It comprises a housing with an air channel for airflow, thin film gas sensors with their active sensor portions positioned in the air channel, and a processor, which may also be configured to perform model selection and deployment.
The processor is configured to supply power to the sensors, collect and process their outputs, and use machine learning algorithms to generate, train, and test a plurality of models. These models are evaluated using a set evaluation parameter, such as prediction accuracy, and a model is selected for deployment based on comparison of this parameter across all models. The sensor data is correlated with specific analyte types, concentrations, or age and quality indicators of food or beverage items, allowing for real-time prediction and notification capabilities.
The problem addressed by this invention is the need for accurate and interpretable identification, classification, and quantification of aromas or analytes for various real-world applications, including the assessment of food freshness, beverage quality, or detection of hazardous chemicals. Traditional methods struggled to precisely and efficiently convert sensor outputs into information that could be intuitively understood or acted upon by users, especially in the context of complex aroma mixtures and patterns.
Claims Coverage
The patent includes two independent claims, each defining a system for determining the age and/or quality of food or beverage. The following inventive features are extracted from the independent claims.
System for determining age and/or quality of food or beverage using an e-nose and machine learning
A system comprising: - An electronic nose (e-nose) with a housing having openings at both ends to enable airflow and an air channel. - A plurality of thin film gas sensors, each with an active sensor portion in the air channel. - At least one identification scanner for reading food or beverage codes, a touch panel for user input, or an image processor for analyzing images to identify the item. - A processor configured to: - Supply power and receive output from the gas sensors. - Predict the age and/or quality of the food or beverage using combinations of sensor outputs and a deployed machine learning model. - Issue a notification of the determination. - Generate random datasets for training and testing multiple models using the received output. - Train and test plural models with various machine learning techniques, evaluating prediction accuracy and deploying a selected model based on comparison of an evaluation parameter.
System for determining age and/or quality based on images and machine learning
A system comprising: - An electronic nose (e-nose) with a housing having airflow openings and an air channel, and a plurality of thin film gas sensors in the air channel. - At least one identification scanner, touch panel, or image processor for food or beverage identification. - A processor configured to: - Power and receive outputs from the sensors. - Predict age and/or quality using combinations of sensor outputs and a deployed machine learning model. - Issue a notification of the determination. - Where the image processor receives images of the food or beverage from different times (including baseline, expired, and spoiled conditions) and the processor determines age based on a deployed machine learning model trained from the images.
These inventive features define systems integrating multi-sensor e-nose hardware, item identification interfaces, and machine learning-based prediction models for determining the age and/or quality of food or beverage, including the use of both sensor output and image data.
Stated Advantages
The system enables accurate prediction of age or quality of food or beverage using machine learning models trained on sensor data, which addresses the complexity of aroma mixtures and enhances determination precision.
Integration of identification scanner, touch panel, or image processor allows for flexible and user-friendly identification of items, facilitating a more intuitive experience for non-technical users.
Automated model selection based on evaluation parameters ensures that the deployed machine learning model is optimized for prediction accuracy.
Documented Applications
Determining the age and/or quality of food or beverage items, including predicting expiration and spoilage conditions using aroma patterns sensed by the electronic nose.
Using images of food or beverage items to augment or replace sensor data for determining age or quality via machine learning models, including image acquisition at various stages (baseline, expired, spoiled).
Interested in licensing this patent?