The newly engineered optical feature extraction engine, known as OFE2, uses an integrated data preparation module to split input signals into parallel optical pathways, ensuring phase stability and minimal delay. This module leverages tunable splitters and precise on-chip delay lines, overcoming the limitations caused by fiber-based phase fluctuations in traditional setups.
OFE2 processes data through an optical diffraction operator, effectively performing matrix-vector multiplication. The resulting focused output light enables precise feature analysis by adjusting the phase of incoming optical signals, giving the system control over output dynamics in real time.
With a demonstrated operation at 12.5 GHz, OFE2 completes single matrix-vector computations in under 250.5 picoseconds, significantly reducing latency compared to electronically based alternatives. Professor Chen noted, "We firmly believe this work provides a significant benchmark for advancing integrated optical diffraction computing to exceed a 10 GHz rate in real-world applications."
The technology was tested in applications including image processing, where it generated enhanced feature maps that improved image classification and semantic segmentation, and in digital trading, where the system directly produced trading signals in response to real-time market input. The system's high speed allows traders to act with minimal delay, supporting stable profits.
This research marks a notable step in shifting computational workloads from energy-consuming electronics to photonic systems, offering potential for real-time decision making in image recognition, healthcare, and finance. Professor Chen concluded, "The advancements presented in our study push integrated diffraction operators to a higher rate, providing support for compute-intensive services in areas such as image recognition, assisted healthcare, and digital finance. We look forward to collaborating with partners who have data-intensive computational needs."
Research Report:High-speed and low-latency optical feature extraction engine based on diffraction operators
Related Links
SPIE--International Society for Optics and Photonics
Stellar Chemistry, The Universe And All Within It
| Subscribe Free To Our Daily Newsletters |
| Subscribe Free To Our Daily Newsletters |
