Real-Time Steering Vector Estimation for Source Localization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional source localization methods using steering vector estimation require priori information about microphone arrangements, leading to inaccurate estimates and noise persistence or distortion, especially in real-time processing scenarios.
Innovation Solution
A real-time steering vector estimation method based on an online complex Gaussian mixture model using a recursive least squares technique, which updates parameters in each time frame without requiring priori information, allowing for accurate source localization and noise removal.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If conventional source localization methods use priori information about microphone arrangements for steering vector estimation, then the estimation process is simplified, but the accuracy deteriorates due to discrepancies between priori information and real microphone locations
Solution Approach 1:
The system performs self-calibration by automatically estimating microphone locations from the input signals themselves without requiring external priori information. The EM algorithm enables the system to self-determine the steering vectors and microphone positions based on the statistical properties of the received signals, making the system self-sufficient and eliminating the need for manual calibration or accurate priori information.
Solution Approach 2:
The invention changes the approach from using fixed priori information parameters to dynamically estimating parameters (steering vectors and microphone locations) from the signals. By treating microphone locations as unknown parameters to be estimated rather than known inputs, the system adapts to the actual physical configuration and achieves accurate localization without relying on potentially inaccurate priori information.
2Measurement precision
If conventional CGMM-based steering vector estimation is used to avoid priori information requirements, then estimation accuracy improves, but real-time processing capability deteriorates due to computational complexity
Solution Approach 1:
The invention introduces dynamic adaptation by recursively updating the steering vectors and covariance matrices at each time frame using the EM algorithm. This allows the system to track time-varying source positions and adapt to changing acoustic environments in real-time, transforming a static estimation approach into a dynamic one that maintains accuracy while enabling real-time processing.
Solution Approach 2:
The invention segments the estimation process into independent time frame iterations, where the EM algorithm is applied recursively to each time frame. By dividing the continuous signal processing into discrete time frames and updating parameters incrementally, the computationally intensive CGMM estimation becomes feasible for real-time applications while maintaining high accuracy.
3Productivity
If recursive least squares technique with online complex Gaussian mixture model is used, then real-time processing capability improves, but computational complexity increases
Solution Approach 1:
The invention maintains continuous useful action by recursively updating the steering vectors and covariance matrices at each time frame without restarting the estimation process. The EM algorithm continues to refine the parameter estimates incrementally, ensuring that the system continuously adapts to new information while maintaining real-time performance through efficient iterative updates rather than repeated full estimations.
Data Source
AI summary
Provided is a source localization in an apparatus for performing a source localization, a target sound source enhancement or speech recognition. The source localization method using input signals input from a plurality of microphones, comprising steps of: (a) obtaining a log likelihood function or an auxiliary function under the assumption that a target source signal mixed with noises satisfies a CGMM model; (b) obtaining an equation for estimating parameter values of the log likelihood function or the auxiliary function so that a value of the log likelihood function or the auxiliary function is maximized recursively in each time frame; (c) estimating a covariance matrix recursively in each time frame; and (d) estimating a steering vector recursively by using the estimated covariance matrix, wherein the steering vector of the target sound source is estimated from the input signals.