.png){kind=icon}
.png){kind=icon}
Return
Integrated Photonic Processor Implementing Digital Image Convolution
.png){kind=icon}
OA
AI
DOI：10.3390/electronics14040709.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Upon the advent of the big data era, information processing hardware platforms have undergone explosive development, facilitating unprecedented computational capabilities while significantly reducing energy consumption. However, conventional electronic computing hardware, despite significant upgrades in architecture optimization and chip scaling, still faces fundamental limitations in speed and energy efficiency due to Joule heating, electromagnetic crosstalk, and capacitance. A new type of information processing hardware is urgently needed for emerging data-intensive applications such as face identification, target tracking, and autonomous driving. Recently, integrated photonics computing architecture, which possesses remarkable compactness, wide bandwidth, low latency, and inherent parallelism, has harvested great attention due to its enormous potential to accelerate parallel data processing, such as digital image convolution. In this study, an integrated photonic processor based on a Mach-Zehnder interferometer (MZI) network is proposed and demonstrated. The processor, being scalable and compatible with complementary metal oxide semiconductors, facilitates mass production and seamless integration with other silicon-based optoelectronic devices. An experimental verification for digital image convolution is also performed, and the result deviations between our processor and a commercial 64-bit computer are less than 2.3%.
Keywords:
photonics computing
optoelectronic integration
artificial intelligence
optical neural networks
integrated photonics
digital image convolution
Journal
IF:
2.6
Papers:
9.5K
Citations:
4.7W