Binaural Activity Pattern Filtering for Cochlear Implant Sound Source Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Cochlear implant users face challenges in distinguishing between useful and interfering sound sources, particularly in environments with multiple sound sources, leading to decreased speech intelligibility.
Innovation Solution
A device and method that utilize two hearing models, one for each ear, to generate a filtered activity pattern by identifying and separating trajectories associated with useful sound sources from those of interfering sources based on time shifts and spatial positioning, effectively suppressing or removing interfering sounds.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If cochlear implants are used to stimulate auditory nerves directly, then hearing capability is improved for patients with inner ear damage, but speech intelligibility decreases significantly in environments with multiple sound sources
Solution Approach 1:
The patent segments the auditory signal processing by creating separate processing channels for different ears (binaural processing). Each ear's activity pattern is processed independently to identify and separate trajectories from different sound sources, allowing the system to selectively enhance speech while suppressing interfering sounds, thereby maintaining speech intelligibility in multi-source environments.
2Measurement precision
If binaural processing with two hearing models is implemented, then sound source separation is improved, but device complexity increases
Solution Approach 1:
The patent creates a simplified computational model of the auditory system that copies essential binaural processing functions. By implementing activity pattern generation and trajectory identification algorithms that mimic neural processing, the system achieves accurate sound source separation without requiring complex hardware modifications, thus balancing measurement precision with device complexity.
Data Source
Figure 1
Figure 2
Figure 3A
AI summary
A device for producing a filtered activity pattern (146), which is based on a first activity pattern (110) on a hearing model of a first ear and a second activity pattern (112) on a hearing model of a second ear, comprises an identifier (120) for identifying a first trajectory in the first activity pattern and a second trajectory in the second activity pattern, which are associated with the same sound event. The device (100) furthermore comprises a determinator (130) for determining whether the two trajectories are associated with a sound event of a wanted sound source. The device furthermore comprises a filter (140) for filtering the first activity pattern or the second activity pattern based on a result of the determinator as to whether a trajectory is associated with a sound event of the wanted sound source, so that in the filtered activity pattern activity events that are associated with a sound event of the wanted sound source dominate, or that the activity events which are not associated with a sound event of the wanted sound source are no longer present in the filtered activity pattern. The described device can additionally be used to implement a source separator. The device according to the invention enables a determination of a filter activity pattern in an efficient, reliable and hearing-adjusted manner, so that in the filtered activity pattern a wanted sound source is emphasized, while interference sound sources are suppressed or damped.