A closed-loop compressive-sensing-based neural recording system.

Area of Science:

• Neuroscience
• Electrical Engineering
• Biomedical Engineering

Background:

• Implantable neural recording devices require efficient data transmission to minimize power consumption.
• Reducing data bandwidth is crucial for large-scale neural recording systems.

Purpose of the Study:

• To describe a low-power, closed-loop compressive sensing (CS) based neural recording system.
• To demonstrate a scalable solution for reducing power consumption in neural recording devices.

Main Methods:

• An application-specific integrated circuit (ASIC) with 4 recording channels utilizing CS circuits was developed.
• A real-time, off-chip CS recovery block with adaptive compression rate adjustment was implemented.
• In vivo and public neural databases were used for system testing.

Main Results:

• Achieved >10x data compression on the neural spike band (500-6KHz) with minimal additional power consumption (0.83uW/electrode).
• Further compression of detected spikes by ~16x was possible, maintaining >95% spike classification success rate.
• The CS circuit occupied 0.11mm²/electrode in a 180nm CMOS process, with total power consumption <16uW/electrode.

Conclusions:

• The system's power and area efficiency make it suitable for large-scale neural recording arrays.
• Closed-loop operation and robust reconstruction enhance the practicality of CS for long-term neural recording.