In-Sensor Optoelectronic Deep Reservoir Computing with Chitosan-Gated Transistors for Action Recognition
Chinese Physics B · 04.06.2026
DOI: 10.1088/1674-1056/ae77cd ↗Kurz gefasst
Fachartikel in der Zeitschrift Chinese Physics B, online erschienen am 04.06.2026. Verfasst von Hangyuan Cui und 4 weiteren Autor:innen.
Abstract (Originalsprache)
Physical reservoir computing (PRC) offers energy-efficient temporal data processing, yet traditional architectures are bottlenecked by the digital overhead of time-division multiplexing and mask matrices. Here, we demonstrate a mask-free deep reservoir computing (deep RC) architecture exploiting the heterogeneous optoelectronic dynamics of chitosan-gated electric double layer (EDL) transistors. Distinct from purely algorithmic models, the network depth is established by directly cascading these physical relaxation processes. Via linear state mapping, this in-materia framework conceptually bypasses multiplexing circuitry, inherently synergizing optical perception with electrical computation for multiscale feature extraction. Validated on a dynamic machine vision task, the system achieves a baseline 90.22% recognition accuracy across 15 human actions, which climbs to 93.78% in the optoelectronic cascading mode, establishing a highly efficient foundation for in-sensor computing.
Zitieren
APA
Cui, H., Yang, Y., Han, M., Shi, K., & Wan, C. (2026). In-Sensor Optoelectronic Deep Reservoir Computing with Chitosan-Gated Transistors for Action Recognition. Chinese Physics B. https://doi.org/10.1088/1674-1056/ae77cd
Vancouver
Cui H, Yang Y, Han M, Shi K, Wan C. In-Sensor Optoelectronic Deep Reservoir Computing with Chitosan-Gated Transistors for Action Recognition. Chinese Physics B. 2026. doi:10.1088/1674-1056/ae77cd
BibTeX
@article{Cui2026,
author = {Cui, Hangyuan and Yang, Yang and Han, Mengtao and Shi, Kailu and Wan, Changjin},
title = {In-Sensor Optoelectronic Deep Reservoir Computing with Chitosan-Gated Transistors for Action Recognition},
journal = {Chinese Physics B},
year = {2026},
doi = {10.1088/1674-1056/ae77cd}
}
EUVIMED