Target Source Signal Generation Using Optimized Separation Filters
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Independent vector extraction methods require increased processing time as the number of microphones increases, hindering efficient sound source extraction.
Innovation Solution
A sound source signal generation device utilizing an optimization algorithm with a separation matrix and auxiliary functions to optimize separation filters, enabling high-speed sound source extraction through algorithms like majorization-minimization and linear constrained minimum variance beamforming.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If independent vector extraction is used for sound source extraction, then sound source separation accuracy is improved, but processing time increases as the number of microphones increases
Solution Approach 1:
The patent segments the sound source extraction process into two distinct phases: an offline preparation phase where the separation matrix is pre-computed using training data, and an online extraction phase where the pre-computed matrix is applied to separate sound sources in real-time. This segmentation allows complex computations to be performed only once during offline processing, while online processing uses the pre-computed results for fast sound source separation, thus resolving the contradiction between separation accuracy and processing time.
Solution Approach 2:
The patent performs preliminary computation of the separation matrix during an offline training phase using training signals from multiple microphones. The separation matrix, which contains the critical information for sound source separation, is pre-calculated and stored. During online sound source extraction, this pre-computed matrix is directly applied without requiring re-computation, enabling fast processing while maintaining high separation accuracy that would otherwise require extensive real-time calculations.
2Measurement precision
If the number of microphones is increased to improve sound source extraction accuracy, then measurement precision is improved, but processing complexity and time increase
Solution Approach 1:
The patent divides the processing into offline matrix computation (which handles the complexity of multiple microphones) and online signal application (which is computationally simple). The separation matrix computed offline encapsulates all the complexity related to the number of microphones, allowing online processing to remain simple regardless of how many microphones are used, thus resolving the contradiction between extraction accuracy and processing complexity.
Solution Approach 2:
The patent creates a computational model (the separation matrix) during offline training that captures the spatial relationships and characteristics of sound sources across multiple microphones. This matrix serves as a copied representation of the complex multi-microphone environment, allowing the system to handle multiple microphones during online processing without repeating the complex computations, thereby reducing processing complexity while maintaining the benefits of using multiple microphones for accurate sound source extraction.
Data Source
AI summary
A sound source signal generation technology based on an optimization algorithm that enables high-speed processing of sound source extraction is provided. A sound source signal generation device includes an optimization unit that optimizes a separation matrix W(f)=[w1(f), . . . , wK(f), WZ(f)] using an observed signal x(f, t), the optimization unit includes an auxiliary function calculation unit that calculates an auxiliary function Vi(f) (i=1, . . . , K) according to a predetermined equation, a first separation filter calculation unit that calculates a separation filters wi(f) (i=1, . . . , K) using auxiliary functions Vi(f) (i=1, . . . , K) and Vz(f), and a second separation filter calculation unit that calculates a separation filter WZ(f) according to a predetermined equation when a convergence condition is satisfied.