Sound Source Localization via Reflection Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Electronic devices face challenges in accurately isolating desired speech from undesired speech and noise sources during voice command processing, as existing sound source localization methods fail to distinguish between direct and reflected sound sources caused by acoustically reflective surfaces.
Innovation Solution
The system performs reflection classification by generating sound track data for each sound source, determining correlation between sound tracks, and using a trained reflection classifier to differentiate between direct and reflected sound sources, thereby ignoring reflections during sound source localization processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sound source localization is performed without reflection classification, then processing is simpler and faster, but accuracy deteriorates due to inability to distinguish direct sound from reflected sound
Solution Approach 1:
The patent segments the sound source identification process into two distinct stages: first, sound tracks are generated from audio data; second, a reflection classifier analyzes correlations between sound tracks to distinguish direct sound sources from reflections. This segmentation allows the system to maintain simplicity in the initial sound tracking while adding complexity only where needed for accurate differentiation, thereby improving localization accuracy without overwhelming computational burden.
Solution Approach 2:
The patent introduces an intermediary reflection classifier component that acts as a mediator between raw audio data and final sound source localization results. This classifier computes correlation values between sound tracks and uses these correlations to identify which tracks represent direct sound versus reflections, serving as an intermediate processing layer that enhances accuracy without requiring complete redesign of the entire localization system.
2Reliability
If all sound tracks are processed equally for localization, then processing is more straightforward, but harmful reflections are not filtered out reducing overall system performance
Solution Approach 1:
The patent implements a feedback mechanism where the reflection classifier continuously analyzes correlation values between sound tracks and uses this information to dynamically adjust which sound tracks are considered valid direct sound sources. The classifier's output feeds back into the sound source localization process, allowing the system to adaptively filter out reflections based on real-time correlation analysis, thereby improving reliability without requiring complex manual configuration.
Solution Approach 2:
The patent changes the parameter of sound track evaluation from simple presence detection to correlation-based differentiation. By computing correlation values between different sound tracks and comparing them against thresholds, the system transforms the detection problem into a parameter-based classification task, making it easier to distinguish direct sound from reflections through quantitative analysis rather than qualitative judgment.
Data Source
AI summary
A system configured to perform sound source localization (SSL) using reflection classification is provided. A device processes audio data representing sounds from multiple sound sources to generate sound track data that includes an individual sound track for each of the sound sources. To detect reflections, the device determines whether a pair of sound tracks are strongly correlated. For example, the device may calculate a correlation value for each pairwise combination of the sound tracks and determine whether the correlation value exceeds a threshold value. When the correlation value exceeds the threshold, the device invokes a reflection classifier trained to distinguish between direct sound sources and reflected sound sources. For example, the device extracts feature data from the pair of sound tracks and processes the feature data using a trained model to determine which of the sound tracks corresponds to the direct sound source.