Taking NIRS-BCIs outside the lab: towards achieving robustness against environment noise

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Near-infrared spectroscopy (NIRS) measures brain activity but is sensitive to environmental noise.
• Developing robust brain-computer interfaces (BCIs) for real-world use requires addressing performance degradation from distractions.
• Auditory distractions and environmental noise can significantly impact NIRS signal quality and mental state classification accuracy.

Purpose of the Study:

• To investigate the impact of environmental noise on NIRS-based mental state classification.
• To develop and evaluate strategies for enhancing the robustness of NIRS-based BCIs against noise.
• To explore the utility of autonomic nervous system (ANS) signals as complementary data for noise compensation.

Main Methods:

• Mental state classification using near-infrared spectroscopy (NIRS) signals from the prefrontal cortex.
• Utilized a hidden Markov model-based classifier to assess performance in silent versus noisy conditions.
• Implemented a hybrid system incorporating autonomic nervous system (ANS) physiological signals (electrodermal activity, skin temperature, blood volume pulse, respiration) and an acoustic monitoring technique for noise event detection.

Main Results:

• NIRS-based classification performance dropped to chance levels in noisy environments compared to silent conditions.
• A hybrid ANS-NIRS system with a startle noise compensation strategy significantly improved classification performance.
• The proposed strategies successfully recovered performance comparable to that achieved in silent conditions.

Conclusions:

• Environmental noise poses a significant challenge for NIRS-based BCIs.
• Integrating ANS signals and acoustic monitoring offers a viable approach to mitigate noise interference.
• The developed hybrid system demonstrates potential for real-world applications of NIRS-based BCIs.