Binaural Hearing Aid Noise Estimation via Segmented Microphone Modes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current noise estimation methods in hearing aids are inadequate for accurately suppressing background noise without affecting the desired signal, particularly due to unreliable voice activity detection and inability to track noise changes during speech activity, leading to unsatisfactory noise suppression or the production of 'musical tones'.
Innovation Solution
A device and method for estimating background noise using binaurally interconnected hearing aids with omnidirectional microphones forming directional microphones with monaural and binaural characteristics, where the output signals from these microphones are linked to enhance noise estimation, allowing for better and more robust noise suppression.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If speech activity detection is used to estimate noise power, then noise estimation can be performed during non-speech periods, but reliable detection of voice activity is difficult and noise tracking during simultaneous voice activity is not possible
Solution Approach 1:
The patent segments the noise estimation problem into two distinct operational modes: monaural noise estimation for non-speech periods and binaural noise estimation for speech periods. This segmentation allows each mode to be optimized independently, with monaural estimation providing reliable baseline noise measurement and binaural estimation enabling continuous tracking during speech activity by exploiting inter-ear time differences.
Solution Approach 2:
The system dynamically switches between monaural and binaural noise estimation modes based on detected speech activity. During non-speech periods, monaural estimation is used; during speech periods, binaural estimation takes over. This dynamic adaptation resolves the contradiction by ensuring the appropriate estimation method is active at the right time, maintaining both reliability and accuracy across different operational conditions.
2Measurement precision
If minimum tracking method is used during speech activity, then background noise can be tracked during speech, but contradictory requirements for smoothing time constants and window length arise that can only be solved on average
Solution Approach 1:
The patent transitions from monaural to binaural noise estimation, adding the dimensional aspect of inter-ear time differences. This additional dimension provides more information for noise tracking during speech activity, allowing the system to resolve noise estimates more accurately without relying on average compromises of smoothing parameters. The binaural approach exploits temporal fine structure differences between ears to achieve more precise tracking.
Solution Approach 2:
The system uses feedback from the detection of speech activity to control the switching between estimation modes. When speech is detected, the system activates binaural estimation with appropriate smoothing parameters; when speech is absent, it uses monaural estimation. This feedback mechanism allows optimal parameter selection based on actual operational conditions rather than average compromises.
3Measurement precision
If speech is present, background noise estimation should be adjusted slowly to avoid classifying speech components as noise, but if no speech is present, interference power estimate should follow temporal fine structure without delay
Solution Approach 1:
The patent implements dynamic adaptation of estimation methodology based on speech presence detection. During speech periods, binaural estimation with slower adaptation is used to avoid speech degradation; during non-speech periods, faster tracking is employed to follow temporal noise variations. This dynamic switching resolves the contradiction by adapting the estimation behavior to current operational conditions.
Solution Approach 2:
The system performs preliminary speech activity detection before committing to a specific noise estimation approach. This preliminary action allows the system to prepare the appropriate estimation mode in advance, ensuring that the correct smoothing and tracking parameters are active before speech or non-speech conditions begin, thereby avoiding delays in temporal response.
4Reliability
If directional microphones are used to suppress noise from specific directions, then speech intelligibility improves in situations with signals from different directions, but the system becomes more complex
Solution Approach 1:
The patent segments the microphone system into two functional components: omnidirectional microphones for capturing all sounds equally, and directional processing for noise suppression. This segmentation allows the system to maintain simple omnidirectional hardware while achieving directional noise suppression through signal processing, resolving the contradiction between performance and hardware complexity.
Solution Approach 2:
The patent introduces signal processing algorithms as an intermediary between the omnidirectional microphones and the output. This intermediary performs the directional noise suppression function without requiring physically directional microphones, thereby achieving speech intelligibility improvement while maintaining simple omnidirectional hardware architecture.
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
The apparatus has hearing devices (1A, 1B) e.g. behind-the-ear hearing devices, including omni-directional microphones (3A, 3B) that are electrically connected with each other to form directional microphones having monaural anti-cardioid characteristics (11), respectively. The microphones of the respective devices are wirelessly connected together to form a directional microphone having a binaural figure of eight characteristic (12). Levels of output signals of the directional microphones having the monaural and binaural characteristics are combined together, respectively. An independent claim is also included for a method for background noise estimation.