Hearing Device Neural Network Voice Activity Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Hearing device systems alter the perception of the wearer's own voice, making it seem unusual and alien, which is undesirable, and existing own-voice detection methods are not effectively adapted for universal use without personalization.
Innovation Solution
A method and system using a neural network-based signal processing apparatus with a filter bank and input stage to detect the activity of the wearer's voice, generating binary detection results without personalization, employing static or adaptive filters for spatial separation and frequency channel processing, and utilizing a binaural hearing device system for improved detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If hearing device systems process and amplify acoustic signals to assist hearing, then hearing capability is improved, but the perception of the wearer's own voice becomes altered and unnatural
Solution Approach 1:
The patent extracts the own-voice signal from the mixed acoustic input by using microphone signals that directly capture the wearer's voice before it is processed and replayed through the hearing device. This extracted own-voice signal is then used to adaptively adjust processing parameters, separating the own-voice component from other environmental sounds to maintain natural perception while still providing hearing assistance.
Solution Approach 2:
The system dynamically changes processing parameters such as gain, compression ratios, and signal filtering based on the detected own-voice activity. When own-voice is detected, the system adjusts parameters to minimize distortion and maintain natural sound quality; when own-voice is absent, parameters are optimized for general hearing enhancement.
2Measurement precision
If existing own-voice detection methods are used, then voice activity can be detected, but the detection is not effectively adapted for universal use without personalization
Solution Approach 1:
The patent creates a universal own-voice detection method that works across different wearers without requiring personalized calibration. The system uses generic acoustic features and statistical models that are applicable to all users, enabling the hearing device to detect and process own-voice signals effectively for any wearer immediately upon use, eliminating the need for individualized setup procedures.
Data Source
AI summary
A method for detecting activity of the own voice of a wearer of a hearing device by way of a signal processing apparatus of the hearing device. A first input signal is generated by a first input transducer, and a second input signal is generated by a second input transducer. The two input signals are supplied to a detection unit of the signal processing apparatus, which has a neural network and an input stage, which is connected in front of the neural network. Information signals are generated by the input stage on the basis of the two input signals and the information signals are evaluated by the neural network. A detection result is output by the detection unit based on the evaluation of the information signals by the neural network.