Fig. 6: Object interactions informed by touch contexts.

a A contextual object interaction framework determined by touch state and touch position. Contains photos by New Africa via Freepik License. b Machine learning assisted data acquisition, feature extraction, and sequence learning to integrate and classify proprioceptive and exteroceptive signals. CNN convolutional neural network; LSTM long short-term memory; Cov convolutional layer) (c) Method for object recognition using proprioceptive feel, and the t- SNE mapping of recognition signals in a feature space. d Classification enhancement with feature extraction. e Dynamic recognition process involving phases of initial contact, proprioceptive alignment, and comprehension during object recognition. f Context-switching drum performance application, where drum classes are switched by changing the holding posture of the drumstick, and playing is controlled by recognizing different wrist motion patterns. Contains icons by macrovector via Freepik License. g Confusion matrices comparing fusion learning and context boundary method for similar wrist motion classification. h Signal analysis of the playing process, where drum switching and playing are controlled by separate signal channels. Contains photos by isometricworld and rafael-19santos via Freepik License.