Presentation Information
[P014]Development of an AI-based scheme for visualizing flavors from aroma compound profiles
*Ryoki Tsuji1,3, Chisa Kato2, Toyomi Ito3, Yuko Terada1,2, Keisuke Ito1,2,3 (1. Guraduate School of Univ. of Shizuoka, 2. Univ. of Shizuoka, 3. DigSense, LLC)
Keywords:
flavor,time series,visualization,animation,AI
"Purpose"
Flavor is a key contributor to food palatability, and its understanding and communication are crucial across the food industry. We have developed AI technologies including F-index CompSM, which predicts verbal flavor descriptors (characters) from aroma compound profiles rapidly, easily, and objectively; F-index PairingSM, which predicts flavor compatibility; and F-index MappingSM, which visualizes flavor similarity. These systems support flavor analysis through language. Nonetheless, as humans process approximately 80% of sensory information visually, applying visual representations may further enhance flavor understanding. Therefore, in this study, we developed a novel system, F-index MovieSM, to visualize time-series changes in flavor as video using the above AIs.
"Methods & Results"
For 399 characters defined by F-index CompSM, five images per character were generated using DALL-E3. ChatGPT was used to evaluate character-image consistency, and 173 images with high agreement were selected. From 8,720 survey responses collected from 97 participants, 112 character-image pairs were validated as perceptually consistent. These were used to visualize the time-series change in retronasal aroma during banana ripening. PTR-MS-based time-series aroma data were converted to characters using F-index CompSM and mapped to corresponding images. Based on threshold values, each image was displayed along a timeline to produce a video of the flavor transition.
This scheme offers a new approach for enhancing the understanding and communication of flavor through visual means.
Flavor is a key contributor to food palatability, and its understanding and communication are crucial across the food industry. We have developed AI technologies including F-index CompSM, which predicts verbal flavor descriptors (characters) from aroma compound profiles rapidly, easily, and objectively; F-index PairingSM, which predicts flavor compatibility; and F-index MappingSM, which visualizes flavor similarity. These systems support flavor analysis through language. Nonetheless, as humans process approximately 80% of sensory information visually, applying visual representations may further enhance flavor understanding. Therefore, in this study, we developed a novel system, F-index MovieSM, to visualize time-series changes in flavor as video using the above AIs.
"Methods & Results"
For 399 characters defined by F-index CompSM, five images per character were generated using DALL-E3. ChatGPT was used to evaluate character-image consistency, and 173 images with high agreement were selected. From 8,720 survey responses collected from 97 participants, 112 character-image pairs were validated as perceptually consistent. These were used to visualize the time-series change in retronasal aroma during banana ripening. PTR-MS-based time-series aroma data were converted to characters using F-index CompSM and mapped to corresponding images. Based on threshold values, each image was displayed along a timeline to produce a video of the flavor transition.
This scheme offers a new approach for enhancing the understanding and communication of flavor through visual means.
Comment
To browse or post comments, you must log in.Log in