Multimodal In-Sensor Computing System Using Integrated Silicon Photonic Convolutional Processor
Author Information
Author(s): Xiao Zian, Ren Zhihao, Zhuge Yangyang, Zhang Zixuan, Zhou Jingkai, Xu Siyu, Xu Cheng, Dong Bowei, Lee Chengkuo
Primary Institution: National University of Singapore
Hypothesis
Can a photonic multimodal in-sensor computing system enhance data processing capabilities for complex spectroscopic sensory data?
Conclusion
The study demonstrates a photonic multimodal in-sensor computing system that achieves a classification accuracy of 97.85% for protein species using integrated silicon photonic technology.
Supporting Evidence
- The system achieved a classification accuracy of 97.85% across 45 different classes of protein species.
- The photonic processor performed convolutional operations with 5-bit accuracy.
- The multimodal in-sensor computing system reduces communication power consumption by processing data at the edge.
- Integration of photonic sensors and processors enhances data processing capabilities.
- The study addresses high communication costs associated with traditional AIoT systems.
Takeaway
This study shows how a special sensor and processor can work together to quickly and accurately identify different proteins, making it easier to analyze complex data.
Methodology
The study used a photonic multimodal sensor integrated with a silicon photonic convolutional processor to classify protein species based on spectroscopic data.
Limitations
The accuracy of the CNN on the ISPCP was slightly lower than that of traditional electronic computation due to precision limitations of the ISPCP chip.
Statistical Information
P-Value
p<0.05
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website