Abstract
In recent years, wearable devices are increasing needs to process large amounts of bio-signals for digital healthcare. However, conventional rigid device causes damage to biological tissue during long-term wearing. And conventional cloud AI signal processing has problems on network load. Here, this study shows a flexible organic memory with low power consumption toward an imperceptible wearable sensor that efficiently performs onsite AI signal processing. The organic memory in a total thickness of 3 μm comprises an organic semiconductor and electrodes on Parylene substrate, which exhibits bending durability, linearly multi-valued memory, and low writing energy. The memory characteristic variated with connecting load resistances is related with the time constant that is calculated with total resistance and total capacitance in the device. The relationship is important when the organic memory that behaves like synapse is implemented in a circuit of AI computing. Furthermore, the synapse behavior of the organic memory is successfully utilized to measure biological signals. In the future, flexible sensors using in-memory computing can contribute to efficient on-site diagnosis under long-term attachment to the skin or clothing.
