Related Experiment Videos
Adaptive, integrated sensor processing to compensate for drift and uncertainty: a stochastic 'neural' approach
1School of Engineering and Electronics, The University of Edinburgh, Edinburgh, United Kingdom.
IEE Proceedings. Nanobiotechnology
|February 16, 2006
Summary
A novel Continuous Restricted Boltzmann Machine (CRBM) adaptive classifier accurately measures H+ ion concentration. This system overcomes sensor noise and drift for real-time analysis in miniaturized devices like Lab-in-a-Pill.
Area of Science:
- Artificial Intelligence
- Biomedical Engineering
- Sensor Technology
Background:
- Accurate H+ ion concentration measurement is crucial for various applications.
- Existing methods face challenges with sensor noise and drift.
- Miniaturized microsystems require robust, real-time analytical capabilities.
Purpose of the Study:
- To demonstrate an adaptive stochastic classifier using a novel neural architecture.
- To achieve accurate H+ ion concentration measurement despite environmental interference.
- To assess the suitability of the system for miniaturized, real-time applications.
Main Methods:
- Development of a Continuous Restricted Boltzmann Machine (CRBM) based classifier.
- Integration of the CRBM with sensors and signal conditioning circuits.
- On-line training to dynamically adapt to sensor drift and noise.
Main Results:
- High accuracy in measuring and classifying H+ ion concentration was achieved.
- The classifier demonstrated robustness against random noise and sensor drift.
- Successful dynamic adaptation to incomplete and drifting sensor data was observed.
Conclusions:
- The CRBM-based adaptive classifier offers a novel solution for accurate H+ sensing.
- The signal-level sensor fusion scheme is suitable for real-time analysis in microsystems.
- This technology has potential applications in miniaturized systems such as Lab-in-a-Pill (LIAP).