Feature-Based Beam Steering for ASR Source Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automated speech recognition (ASR) systems, such as smart speakers and IoT devices, face challenges in distinguishing desired speech sources from interference due to the inability to exclude known sources effectively, often treating interference as desired signals, which hampers accurate beamforming and speech enhancement.
Innovation Solution
A method that identifies multiple sources by assigning feature values to each source, determining scores based on these features, and selecting sources for spatial processing, incorporating features like source persistency, activity, mobility, and energy to improve beam steering and reject undesired signals, even with a single beamformer.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If beamforming systems select sources based solely on largest activity, then the system is simple to operate, but interference sources are incorrectly treated as desired sources
Solution Approach 1:
The system changes the parameters used for source selection from simple activity level to a composite score based on multiple features including activity, persistency, mobility, and energy. This allows the beamformer to distinguish between desired speech sources and interference sources by evaluating multiple characteristics rather than relying on a single parameter.
Solution Approach 2:
The patent introduces an intermediary scoring mechanism that evaluates multiple features before making beam steering decisions. This intermediary layer processes source characteristics and produces a composite score that guides the beamformer, preventing direct selection based solely on activity level and enabling more accurate source identification.
2Measurement precision
If multiple features are evaluated for each source, then source identification accuracy is improved, but computational complexity increases
Solution Approach 1:
The system segments the source evaluation process into distinct feature extraction steps (activity, persistency, mobility, energy) that can be computed independently and then combined. This modular approach allows for efficient processing by breaking down the complex evaluation into manageable components that can be calculated and integrated systematically.
3Reliability
If feature-based scoring is implemented, then interference rejection is improved, but processing time increases
Solution Approach 1:
The system performs preliminary evaluation of source features (activity, persistency, mobility, energy) before the beamforming decision is made. By pre-computing these features and combining them into a composite score in advance, the system reduces the processing time required during actual beam steering operations, as the scoring framework is already established and ready for rapid decision-making.
Data Source
AI summary
A method, computer program product, and computer system for identifying, by a computing device, a plurality of sources. One or more feature values of a plurality of features may be assigned to a first source of the plurality of sources. One or more feature values of the plurality of features may be assigned to a second source of the plurality of sources. A first score for the first source and a second score for the second source may be determined based upon, at least in part, the one or more feature values assigned to the first source and the second source. One of the first source and the second source may be selected for spatial processing based upon, at least in part, the first score for the first source and the second score for the second source.