Pulsewidth Modulation-Based Algorithm for Spike Phase Encoding and Decoding of Time-Dependent Analog Data

Area of Science:

• Neuroscience
• Signal Processing
• Computer Science

Background:

• Spiking neural networks (SNNs) offer a biologically plausible model for information processing.
• Efficient encoding and decoding of analog data are crucial for SNN performance.
• Current methods face challenges in balancing reconstruction accuracy and data compression.

Purpose of the Study:

• To propose a novel spike encoding and decoding algorithm for analog data.
• To demonstrate the algorithm's effectiveness using benchmark datasets.
• To explore applications in SNN modeling and neuromorphic computing.

Main Methods:

• The algorithm employs pulsewidth modulation principles for spike generation.
• It processes analog data, including stock index time series and human voice data.
• Performance is evaluated based on signal reconstruction accuracy and data compression ratios.

Main Results:

• The proposed algorithm achieves high reconstruction accuracy for analog signals.
• Significant data compression is attained, reducing the data load for SNNs.
• The method is validated on diverse datasets, showing broad applicability.

Conclusions:

• The novel algorithm effectively encodes and decodes analog data for SNNs.
• It enables high-fidelity signal reconstruction and efficient data compression.
• This work facilitates new SNN applications in predictive time-series modeling and neuromorphic hardware.