Binaural Hearing System DOA Estimation via Neural Network Spatial Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting the direction of arrival (DOA) of an acoustic target signal in noisy environments, particularly in binaural hearing systems, face challenges in accuracy and robustness due to limited processing resources and the presence of multiple target sources.
Innovation Solution
A method utilizing a plurality of microphones distributed across a local and remote device in a binaural hearing system, where spatial feature quantities are derived from pairs of input signals and used as input to a neural network to estimate the DOA of the acoustic target signal.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional DOA estimation methods are used in noisy environments with multiple targets, then processing resources are consumed, but detection accuracy deteriorates
Solution Approach 1:
The patent segments the acoustic signal processing by separating target signal extraction from spatial feature extraction. The neural network segments different spatial cues (ITD, ILD, spectral features) into distinct feature quantities. This segmentation allows the system to process only relevant features for DOA estimation, improving accuracy while reducing the impact of background noise and computational complexity.
2Reliability
If complex signal processing is applied to enhance target signal in noisy environment, then signal-to-noise ratio improves, but processing complexity increases
Solution Approach 1:
The patent introduces spatial feature quantities as intermediary representations between raw microphone signals and DOA estimation. These feature quantities (including ITD, ILD, and spectral features) serve as mediators that capture essential spatial information while filtering out noise. The neural network processes these intermediaries rather than raw signals, reducing processing complexity while maintaining reliability.
3Measurement precision
If multiple spatial feature quantities are derived from different microphone signal pairs, then DOA estimation accuracy improves, but computational load increases
Solution Approach 1:
The patent applies preliminary action by pre-defining the set of spatial feature quantities to be extracted from microphone signal pairs before actual DOA estimation. The system pre-processes microphone signals to extract ITD, ILD, and spectral features, organizing them into structured feature quantities. This preliminary organization reduces real-time computational load while maintaining high detection accuracy through comprehensive spatial feature coverage.
Data Source
AI summary
A method detects a direction of arrival of an acoustic target signal. A local device contains first and second local microphones, and a remote device contains a first remote microphone. The method includes the steps of: deriving a first local input signal from first and second local microphone signals, deriving a second local input signal from the first and/or second local microphone signals, and deriving a first remote input signal from the first remote microphone. The first and second local input signals and first remote input signal form a part of a set of input signals. A plurality of spatial feature quantities are each derived from different respective pairs, and are indicative of a spatial relation between the two corresponding input signals. The spatial feature quantities are input to a neural network. The direction of arrival of the acoustic target signal is estimated in the neural network.