Textile In-Memory Computing Memristor for Image Compression Applications
Abstract
Wearable edge computing devices require novel hardware for high-efficiency data processing. In this work, we proposed novel textile in-memory computing memristor, exhibiting great potential in image compression applications for the first time. The device shows low operation voltage and reliable switching characteristics, paving the way for continuous analog modulation of input signals. Based on the experimental I-V curve model, 16-cell memristor array was constructed to validate analog multiplication and integration functions of memristor circuits. The measured current and voltage responses were further obtained from the practical circuit and showed good agreement with the theoretical results. To further evaluate the feasibility of realizing compressed sensing tasks, block-wise compressed sensing reconstruction based on memristor arrays was carried out in practical circuits. These results demonstrate the great potential of textile memristors in constructing next-generation wearable in-memory compressed sensing system.
