Cochlear Implant Frequency Estimation Using Temporal Fine Structure
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current cochlear implant technologies face challenges in accurately estimating instantaneous frequency, especially in the presence of multiple harmonics and noise, leading to poor speech intelligibility and music perception due to limitations in existing zero-crossing techniques and filter bank resolutions.
Innovation Solution
A signal processing arrangement that generates electrode stimulation signals by extracting band pass signals, using a timing function to represent instantaneous frequency based on temporal fine structure features, excluding short-term temporal features, and applying a smoothing window to improve robustness, while preserving interaural time difference information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If zero-crossing techniques are used for instantaneous frequency estimation, then the processing is computationally simple, but the estimation accuracy deteriorates in the presence of multiple harmonics and noise
Solution Approach 1:
The patent introduces an intermediary model (sum of sinusoids with time-varying amplitude and frequency) between the raw signal and the frequency estimation. This model acts as a mediator that separates the fundamental frequency component from harmonics and noise, allowing accurate tracking of instantaneous frequency without being affected by multiple harmonics. The model parameters are estimated using an iterative algorithm that refines the frequency, amplitude, and phase values.
Solution Approach 2:
The patent changes the parameter representation from fixed zero-crossing points to time-varying model parameters (frequency, amplitude, phase). By modeling the signal as having time-varying parameters rather than relying on fixed zero-crossing detection, the system can adapt to changing signal conditions and maintain accurate frequency estimation even in the presence of harmonics and noise.
2Measurement precision
If high-resolution filter banks are used to improve frequency estimation, then the measurement precision improves, but the device complexity and processing requirements increase
Solution Approach 1:
The patent extracts the essential frequency information directly from the temporal fine structure of the band-pass signals without requiring high-resolution filter banks. By taking out only the necessary temporal features (zero-crossings, peak detections, or phase information) and processing them through the iterative model, the system achieves accurate frequency estimation while avoiding the complexity of high-resolution filter banks.
Solution Approach 2:
The patent replaces the mechanical/filter-based frequency analysis system with a computational model-based approach. Instead of using complex physical filter banks to resolve frequencies, the system uses an iterative algorithm that computes frequency, amplitude, and phase parameters directly from the signal, substituting mechanical filtering with computational modeling.
3Loss of information
If all temporal fine structure features are used for stimulation timing, then the instantaneous frequency representation is complete, but noise and short-term features degrade the estimation robustness
Solution Approach 1:
The patent performs preliminary selection and validation of temporal fine structure features before using them for stimulation timing. By pre-screening features to ensure they meet robustness criteria (excluding noise-dominated regions and short-term artifacts), the system prepares clean, reliable data for the iterative model, improving the overall robustness of frequency estimation while maintaining information completeness.
Solution Approach 2:
The patent implements feedback mechanisms where the iterative model continuously refines its parameter estimates based on the measured signal characteristics. The model uses feedback from the signal data to adjust frequency, amplitude, and phase estimates iteratively, and also provides feedback to validate whether detected features are reliable, thereby improving robustness while maintaining completeness.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
A signal processing arrangement generates electrical stimulation signals to electrode contacts in an implanted cochlear implant array. An input sound signal is processed to generate band pass signals that each represent an associated band of audio frequencies. A characteristic envelope signal is extracted for each band pass signal based on its amplitude. Stimulation timing signals are generated for each band pass signal, including for one or more selected band pass signals using a timing function defined to: i. represent instantaneous frequency as determined by the band pass signal temporal fine structure features, and ii. exclude temporal fine structure features occurring within a time period shorter than a band-specific upper frequency limit. The electrode stimulation signals are produced for each electrode contact based on the envelope signals and the stimulation timing signals.