An Event-Based Digital Time Difference Encoder Model Implementation for Neuromorphic Systems.

Area of Science:

• Neuromorphic Engineering
• Signal Processing
• Bio-inspired Computing

Background:

• Neuromorphic systems excel in real-time, resource-constrained tasks due to low power and latency.
• Event-based processing reduces computational load and data loss via sparse, adaptive sampling.
• Extracting temporal information from event-based signals using rate coding presents challenges.

Purpose of the Study:

• To introduce a novel digital implementation, the time difference encoder (TDE), for temporal encoding in event-based signals.
• To develop a bio-inspired model that translates time differences between events into encoded output bursts.
• To validate the TDE's efficacy as an alternative to the Jeffress model for interaural time difference estimation.

Main Methods:

• Digital implementation of the time difference encoder (TDE) with a configurable time constant.
• Encoding temporal information by translating time differences between input events into output event bursts.
• Validation using a sound source lateralization proof-of-concept system and simulation on a field-programmable gate array (FPGA).

Main Results:

• The TDE model successfully encodes temporal information through the number and timing of output events.
• The digital circuit implementation demonstrated low hardware resource utilization (122 slice registers) and minimal power consumption (<1 mW).
• Proof-of-concept validation confirmed the TDE's capability for interaural time difference estimation.

Conclusions:

• The TDE provides an effective method for temporal encoding in event-based neuromorphic systems.
• The configurable digital implementation is suitable for diverse sensing tasks requiring precise time difference encoding.
• This bio-inspired approach offers a power-efficient and resource-light alternative for temporal signal processing.