2023年第70回応用物理学会春季学術講演会

講演情報

一般セッション(口頭講演)

3 光・フォトニクス » 3.2 情報フォトニクス・画像工学(旧3.3)

[15p-A202-1~14] 3.2 情報フォトニクス・画像工学(旧3.3)

2023年3月15日(水) 13:00 〜 17:30 A202 (6号館)

竪 直也(九大)、茨田 大輔(宇都宮大)、最田 裕介(和歌山大)

16:45 〜 17:00

[15p-A202-12] [The 53rd Young Scientist Presentation Award Speech] Double phase hologram based high-capacity holographic memory

Jianying Hao1,2、Xiaoqing Zheng1、Xiao Lin1、Ryushi Fujimura2,3、Soki Hirayama2、Yoshito Tanaka2、Xiaodi Tan1、Tsutomu Shimura2 (1.Fujian Normal Univ.、2.The Univ. of Tokyo、3.Utsunomiya Univ.)

キーワード:Holographic Data Storage, Deep Learning, Double Phase Hologram

The storage capacity of holographic memory is determined by the hologram recording area, the number of holograms multiplexing and the encoding efficiency of one data page. Encoding efficiency can be improved by utilizing multi-dimensional modulation such as complex amplitude. In this report, a double phase hologram (DPH) based complex amplitude-modulated holographic memory is proposed. In the encoding process, only one phase spatial light modulator (SLM) is needed to realize the complex amplitude modulation. In the decoding process, using deep learning to demodulate the complex amplitude from the intensity image captured directly by a detector. A convolutional neural network (CNN) is used to detect the features of the intensity distribution to establish the relationship between the captured intensity and the data pages. After training, the corresponding amplitude and phase data can be inferred directly from the intensity of the reconstructed beam through the trained CNN.