Low-power differencing feature extracts spiking-band activities for high-performance intracortical brain-computer...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Intracortical brain-computer interfaces (iBCIs) require efficient feature extraction for high-bandwidth neural signals.
• Resource constraints in implantable systems necessitate computationally lean algorithms.

Purpose of the Study:

• To introduce and validate the Mean Absolute of n-th Difference (MAND) as a computationally efficient feature extraction technique for iBCIs.
• To compare MAND's performance against existing methods in various neural decoding tasks.

Main Methods:

• Developed MAND, a feature extraction method using optimized differencing operations to isolate neural spiking activity.
• Validated MAND theoretically and empirically across human, primate, and rodent neural datasets.
• Implemented an extended MAND variant with dual-differencing for enhanced spectral alignment.

Main Results:

• MAND significantly reduced velocity reconstruction error and improved classification accuracy compared to state-of-the-art features.
• The extended MAND variant further enhanced performance through improved spectral alignment.
• Hardware implementation demonstrated MAND's exceptional efficiency: 6ms processing time and 3mW power consumption for 10s recordings.

Conclusions:

• MAND offers a breakthrough in computational efficiency for neural signal processing in iBCIs.
• The method enables superior decoding performance, paving the way for advanced fully implantable iBCI systems.
• MAND represents a significant advancement in energy-efficient, high-speed neural feature extraction.