Sparse Signal Space Model for Neuronal Action Potential Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting neuronal action potential signals in hearing implant systems, such as cochlear implants, face challenges in effectively separating neural action potentials from artifacts and noise, leading to suboptimal signal processing and detection accuracy.
Innovation Solution
A sparse signal space model is used to map tissue response measurements into a separable signal space with disjoint manifolds for neural action potential, stimulation artifact, and noise components, allowing for robust and computationally efficient detection and classification of neural action potentials using MOD or K-SVD trained models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional signal processing methods are used to detect neural action potentials, then detection capability is achieved, but signal separation accuracy deteriorates due to inability to effectively separate neural action potentials from artifacts and noise
Solution Approach 1:
The patent applies segmentation by decomposing the mixed signal into distinct component manifolds (neural action potential manifold, artifact manifold, and noise manifold) in a sparse signal space. This allows separate processing and identification of each signal component, effectively separating neural action potentials from artifacts and noise through geometric projection operations in the transformed space.
Solution Approach 2:
The patent introduces a sparse signal space transformation as an intermediary step between the raw mixed signal and the final detection result. This transformation space acts as a mediator that separates overlapping signal components into distinct manifolds, enabling accurate identification of neural action potentials that are otherwise obscured in the time domain.
2Measurement precision
If complex signal processing algorithms are used to improve detection accuracy, then measurement precision is improved, but computational complexity increases
Solution Approach 1:
The patent replaces complex iterative signal processing algorithms with a geometric projection approach in sparse signal space. Instead of using computationally intensive methods like adaptive filtering or machine learning classifiers, the solution uses straightforward linear projection operations onto predefined manifolds, significantly reducing computational complexity while maintaining high detection accuracy.
Solution Approach 2:
The patent transforms the signal from the time domain to a sparse signal space representation, changing the parameter space in which processing occurs. This transformation allows simple projection operations to achieve what would require complex algorithms in the original domain, effectively trading representation complexity for operational simplicity.
3Measurement precision
If traditional recording methods are used, then measurement capability is maintained, but signal-to-noise ratio deteriorates due to super-position of neural responses with artifacts and noise
Solution Approach 1:
The patent extracts neural action potential signals from the contaminated mixed signal by projecting the sparse signal representation onto the neural action potential manifold. This extraction process separates the desired neural information from artifacts and noise, recovering the pure signal component while discarding contaminating elements.
Solution Approach 2:
The patent moves the signal analysis from the one-dimensional time domain to a multi-dimensional sparse signal space where different signal components occupy distinct geometric manifolds. This dimensional transformation enables separation of mixed signals that are inseparable in the time domain, effectively increasing the signal-to-noise ratio through geometric decomposition.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A system and method detect neuronal action potential signals from tissue responding to electrical stimulation signals. A sparse signal space model for a set of tissue response recordings has a signal space separable into a plurality of disjoint component manifolds including a neural action potential (NAP) component manifold corresponding to tissue response to electrical stimulation signals. A response measurement module is configured to: i. map a tissue response measurement signal into the sparse signal model space to obtain a corresponding sparse signal representation, ii. project the sparse signal representation onto the NAP component manifold to obtain a sparse NAP component representation, iii. when the sparse NAP component representation is greater than a minimum threshold value, report and recover a detected NAP signal in the tissue response measurement signal.