Researchers from the Massachusetts Institute of Technology in Cambridge, Mass., developed the glove from commercially available materials for around $10. As a hand wearing the glove picks up and feels various objects, 548 tiny sensors sewn into it capture pressure signals, which are then processed by a neural network, allowing the AI system to classify each object and predict its weight without any visual input.
According to the study, when the glove was used to assess 26 common objects such as a soda can, tennis ball, spoon and mug, the system identified the objects with 76 percent accuracy, and correctly predicted the weights of most objects within about 60 grams.
Used in combination with other AI systems and visual datasets, the glove could enable robots to interact with objects with more human-like understanding, the researchers noted. It could also improve the design of human prosthetics to make them more lifelike and precise.
More articles about AI:
Amazon’s cloud service releases machine learning tool for text, data extraction
Samsung, Stanley Black & Decker invest in countertop AI medication reminder tool
Verily’s AI predicts disease biomarkers ‘significantly better’ than existing models
