An On-Chip Processor for Chronic Neurological Disorders Assistance Using Negative Affectivity Classification

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computer Science

Background:

• Chronic neurological disorders (CNDs) like Alzheimer's, ASD, and ALS degrade cognitive and emotional abilities.
• Continuous neuro-feedback monitoring is vital for managing CNDs' severe effects.
• Early intervention through preemptive measures can alleviate CNDs' impact.

Purpose of the Study:

• To present a hardware-efficient, dedicated processor for human emotion classification in CND patients.
• To enable continuous, non-invasive monitoring for better CND management.
• To improve early detection and intervention strategies for CNDs.

Main Methods:

• Utilized scalp electroencephalography (EEG) for emotion classification based on valence and arousal.
• Employed a linear support vector machine (SVM) classifier with power spectral density features.
• Developed a novel look-up-table based logarithmic division unit (LDU) for efficient feature extraction in machine learning (ML).

Main Results:

• Achieved classification accuracies of 72.96% for valence and 73.14% for arousal.
• The implemented LDU reduced integer division cost by 34% for ML applications.
• The 2x3mm² processor fabricated using a 0.18μm CMOS process demonstrated low power (2.04 mW) and energy (16 μJ/classification) consumption.

Conclusions:

• The developed processor offers a viable, hardware-efficient solution for non-invasive emotion classification in CND patients.
• This technology can support continuous monitoring and facilitate early intervention strategies.
• The efficient LDU design contributes to cost reduction in ML applications for wearable systems.