Comparison metrics and power trade-offs for BCI motor decoding circuit design.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computer Engineering

Background:

• Brain-Computer Interfaces (BCIs) are vital for rehabilitation and assistive technologies, requiring low-power hardware for implantable and battery-operated devices.
• Effective motor and speech decoding necessitates efficient signal processing in Brain-Computer Interfaces (BCIs).

Purpose of the Study:

• To review existing hardware systems for BCIs, focusing on motor decoding, and analyze factors affecting power and algorithmic performance.
• To propose metrics for comparing the energy efficiency of on-chip decoding systems for Electroencephalography (EEG), Electrocorticography (ECoG), and Microelectrode Array (MEA) signals.

Main Methods:

• Analysis of existing hardware systems for BCIs, with a focus on motor decoding.
• Development and application of metrics to compare energy efficiency across various on-chip decoding systems (EEG, ECoG, MEA).
• Review of optimizations in state-of-the-art decoding circuits to minimize power consumption.

Main Results:

• An empirically estimable Input Data Rate (IDR) is required for a given classification rate, aiding in the design of new BCI systems.
• A counter-intuitive negative correlation exists between power consumption per channel (PpC) and Information Transfer Rate (ITR).
• Increasing the number of channels can decrease PpC via hardware sharing and increase ITR by incorporating more data; power consumption is dominated by signal processing complexity for EEG and ECoG.

Conclusions:

• Hardware sharing and increased channel count offer a pathway to simultaneously reduce power consumption per channel and enhance information transfer rates in BCIs.
• Optimizing signal processing complexity is key to minimizing power consumption in EEG and ECoG decoding circuits for Brain-Computer Interfaces (BCIs).
• The findings provide valuable insights for designing energy-efficient BCI systems for real-world assistive applications.