Fig. 5: Tactile near-sensor computing for real-time sensing and processing of pressure stimuli, integrating hourglass-shaped microstructured capacitive sensors and flexible memristors.

a Conceptual illustration of the near-sensor computing system for noise reduction in fingerprint stimuli. b System configuration of a 3 × 3 mean filter for processing a single fingerprint pixel. c Schematic representation of processing noisy fingerprint stimuli to yield a denoised signal image in the tactile sensory nerves. For each pixel’s denoising, currents from the 3 × 3 pressure sensors were averaged using the mean filter. d Pixel bit matching accuracy between the original fingerprint image and the noisy or processed signals in the near-sensor computing. e Average energy consumption for denoising 60 × 60 fingerprint signals and a single pixel within the system. f Schematic of the edge-detection process for fingerprint stimuli, utilizing the same 3 × 3 filter structure as in mean filtering. g Fingerprint signal image generated by edge detection based on pressure stimuli detected within the system.