Blind Source Separation with Smoothed Frequency-Bin Gradients
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Convolutive blind source separation methods face challenges in accurately estimating source signals due to permutation issues and the need for improved smoothness of gradient terms across frequency bins, which affects the effectiveness of unmixing filters in audio signal processing.
Innovation Solution
A method involving transforming input signals into the frequency domain, calculating unmixing filter coefficients using a gradient descent process with adjusted gradient terms to enhance smoothness across frequency bins, and estimating source signals through these filters, while adapting the cost function to evaluate decorrelation between estimated signals.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If gradient descent process is used to calculate unmixing filter coefficients, then source signals can be separated, but the gradient terms lack smoothness across frequency bins causing permutation issues
Solution Approach 1:
The patent modifies the gradient terms by applying smoothing operations across frequency bins during the gradient descent process. This involves changing the parameters of the gradient calculation to ensure continuity and smoothness, thereby resolving permutation issues while maintaining separation accuracy.
Solution Approach 2:
The patent introduces a feedback mechanism where the smoothness of gradient terms across frequency bins is continuously monitored and adjusted. By evaluating the permutation consistency and adjusting gradient terms accordingly, the system achieves stable source separation without permutation errors.
2Measurement precision
If unmixing filters are applied to separate source signals, then signal separation is achieved, but computational complexity increases
Solution Approach 1:
The patent divides the frequency spectrum into multiple bins and processes each bin independently through the gradient descent algorithm. This segmentation allows parallel computation and reduces the overall computational complexity while maintaining accurate source signal separation across the entire frequency range.
Solution Approach 2:
The patent applies local optimization by adjusting gradient terms specifically at frequency bin boundaries to ensure smoothness, rather than processing the entire frequency spectrum uniformly. This localized approach reduces computational burden while maintaining separation quality.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
Methods and apparatuses for convolutive blind source separation are described. Each of a plurality of input signals is transformed into frequency domain. Values of coefficients of unmixing filter corresponding to frequency bins are calculated by performing a gradient descent process on a cost function at least dependent on the coefficients of the unmixing filters. In each iteration of the gradient descent process, gradient terms for calculating the values of the same coefficient of the unmixing filters are adjusted to improve smoothness of gradient terms across the frequency bins. With respect to each of the frequency bins, source signals are estimated by filtering the transformed input signals through the respective unmixing filter configured with the calculated values of the coefficients. The estimated source signals on the respective frequency bins are transformed into time domain. The cost function is adapted to evaluate decorrelation between the estimated source signals.