Food texture estimation of gel-type foods by using a vision-based tactile sensor

This study presents a novel method for quantitative estimation of gel-type food texture using a vision-based tactile sensor. A transparent silicone elastomer layer, containing embedded markers and having an apparent elastic modulus similar to that of the human tongue, was used as the sensor’s soft contact surface. Marker displacements were recorded during compression–fracture tests of eight gel samples. Texture features were extracted and correlated with sensory evaluation scores for “smoothness (tsuru-tsuru),” “elasticity (mochi-mochi),” “granularity (zara-zara),” and “stickiness (nettori)” to build regression models in accordance with previously established methods. The use of marker displacement data as input features is a novel contribution of this study. Model accuracy was assessed by leave-one-out cross-validation, yielding high determination coefficients (R² = 0.861–0.928). The findings demonstrated the effectiveness of vision-based tactile sensing for high-resolution evaluation of gel-type food textures, suggesting its potential utility in the development and quality control of texture-modified foods for elderly individuals or patients with swallowing difficulties.