Analog-to-spike encoding and time-efficient RF signal processing with photonic neurons

Area of Science:

• Photonics
• Signal Processing
• Neuroscience

Background:

• Real-time radio-frequency (RF) signal processing is crucial for advanced systems.
• Conventional methods face latency issues due to high-speed analog-to-digital conversion (ADC) and intensive digital computation.

Purpose of the Study:

• To propose and investigate a novel method for encoding and processing RF signals in the analog domain using photonic spiking.
• To analyze the dependence of photonic analog-to-spike encoding on threshold level and time constant.

Main Methods:

• Theoretical and experimental investigation of photonic spiking neuron response.
• Utilizing photonic spiking for RF signal encoding and processing in the analog domain.
• Waveform classification task using the proposed photonic-spiking-based scheme.

Main Results:

• Photonic spiking neuron demonstrated distinct spike number distributions for different RF waveforms.
• Achieved 92% accuracy in waveform classification, comparable to digital K-nearest neighbor (KNN) at 94%.
• Reduced processing latency from 0.7 seconds to 80 nanoseconds, a >1 million-fold improvement.

Conclusions:

• Photonic spiking response offers a viable, ultra-low-latency solution for real-time RF signal processing.
• The asynchronous encoding and binary output of photonic spiking are key advantages.
• This approach could enable next-generation radar and communication systems.