Particle Filtering Sound Source Location Using Transfer Function Correlation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current particle filtering methods for sound source location estimation face challenges in achieving accurate and efficient weight updates, leading to suboptimal results due to the complexity and reliance on measurement techniques that do not accurately reflect the relationship between state variables and real sound source positions.
Innovation Solution
The method generates particle weights based on the correlation between estimated acoustic transfer functions from the sound source to multiple sound recording positions, using this correlation as a pseudo likelihood function to update weights and improve the representation of noise and uncertainty, allowing for more accurate sound source location estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional measurement functions are used to update particle weights, then the particle filtering process can be implemented, but the accuracy of sound source location estimation deteriorates due to suboptimal weight updates
Solution Approach 1:
The patent changes the parameter used for weight update from traditional measurement functions to inter-microphone transfer function correlations. This parameter change enables more accurate sound source location estimation by using correlation values that directly reflect the spatial relationship between microphones and the sound source, thereby improving both measurement precision and weight update reliability
Solution Approach 2:
The patent substitutes traditional measurement-based weight updating with a correlation-based approach using transfer functions. This substitution replaces the mechanical measurement system with an acoustic field-based system that uses the correlation between transfer functions from multiple microphones to determine particle weights, leading to improved estimation accuracy
2Measurement precision
If complex measurement techniques are used to accurately reflect the relationship between state variables and sound source positions, then estimation accuracy improves, but computational complexity increases
Solution Approach 1:
The patent extracts the essential relationship information between state variables and sound source positions by using only the correlation of transfer functions, discarding the complexity of traditional measurement techniques. This extraction approach maintains high estimation accuracy while significantly simplifying the weight updating process by focusing only on the critical correlation parameter
3Productivity
If traditional particle filtering weight updates are used, then the method can be implemented, but efficiency deteriorates due to suboptimal convergence
Solution Approach 1:
The patent implements a feedback mechanism where the correlation between transfer functions provides direct information about particle quality. This feedback enables more efficient weight updates by using correlation values that directly indicate how well each particle represents the actual sound source position, leading to faster convergence and improved estimation efficiency while reducing computational time
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enhances the accuracy and efficiency of sound source location estimation by providing a reliable indication of the sound source position, reducing complexity and improving weight updating processes, particularly suitable for multi-modal audiovisual applications like teleconferencing.
Implementation Method 1
the step of generating the weight for the particle at the time instant comprises determining the weight in response to a correlation between estimated acoustic transfer functions from the sound source to at least two sound recording positions for the first time instant
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
A sound source location is estimated by particle filtering where a set of particles represents a probability density function for a state variable comprising the sound source location. The method includes determining the weight for a particle in response to a correlation between estimated acoustic transfer functions from the sound source to at least two sound recording positions. A weight update function may specifically be determined deterministically from the correlation and thus the correlation may be used as a pseudo- likelihood function for the measurement function of the particle filtering. The acoustic transfer functions may be determined from an audio beamforming towards the sound source. The audio weight may be combined with a video weight to generate a multi-modal particle filtering approach.