Improving Detection Accuracy of Memristor-Based Bio-Signal Sensing Platform

Area of Science:

• Neuroscience
• Materials Science
• Electrical Engineering

Background:

• A novel neuronal activity sensor utilizes memristor devices for their intrinsic thresholded integrator capabilities.
• Memristors store significant neural signal peaks as non-volatile resistive state changes, effectively suppressing noise.

Purpose of the Study:

• To explore signal processing and memristor operating strategies to enhance neuronal activity detection rates.
• To address the limitation of memristor saturation due to accumulated resistive state changes during operation.

Main Methods:

• Analysis of data from a single memristive device using a reference neural recording.
• Implementation of strategies including adjusting memristor reset frequency and preconditioning input waveforms (gain and offset).
• Benchmarking performance against a state-of-the-art template matching system using principal component analysis.

Main Results:

• Performance improvement was achieved by increasing the frequency of memristor resets to a highly responsive initial resistive state.
• Optimal preconditioning of the input waveform, through gain and offset adjustments, significantly enhanced device performance.
• The proposed strategies markedly improved spike detection capabilities compared to baseline operation.

Conclusions:

• Signal processing and optimized operating procedures can significantly improve the detection capabilities of memristor-based neuronal sensors.
• These findings offer a pathway to overcome memristor saturation and enhance the reliability of neural activity monitoring.