Nonlinear Signal Mixture Estimation for Time-Frequency Support Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating radio signals from multiple sources with unknown directional vectors fail to accurately determine the temporal and spectral supports of the signals, leading to incorrect signal processing even when the input is composed of noise.
Innovation Solution
A method using a network of P>2 antennas to estimate the directional vectors and apply a conditional expectation estimator with linear filters, determining the temporal and spectral supports of signals by analyzing time-frequency representations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If linear filtering (MV) or pseudo-linear filtering (Capon) is used for signal estimation, then the signal-to-noise ratio is improved, but the temporal and spectral supports of the signal cannot be correctly delimited
Solution Approach 1:
The patent changes the estimation approach from linear filtering to non-linear estimation by computing conditional expectations. This transforms the estimation parameter from simple linear combinations to probabilistic expectations that can capture signal presence/absence, thereby achieving both noise suppression and accurate support delimitation.
Solution Approach 2:
The patent introduces conditional expectation as an intermediary between the observed data and the signal estimate. This intermediary allows the system to incorporate prior knowledge about signal statistics and achieve more precise support delimitation while maintaining noise robustness.
2Measurement precision
If maximum likelihood estimation is used, then unbiased linear estimate with minimal variance is obtained, but a priori knowledge of the signal is not exploited
Solution Approach 1:
The patent implements feedback by using computed conditional expectations to refine the estimation process. The algorithm iteratively updates estimates based on accumulated information, allowing a priori knowledge to be progressively exploited while maintaining estimation precision.
Solution Approach 2:
The patent performs preliminary computation of conditional expectations using available prior knowledge before final signal reconstruction. This preliminary action allows the system to prepare optimized estimates that incorporate all available information, reducing the need for iterative refinement.
3Productivity
If linear filtering is applied, then processing is simple and fast, but the output signal is provided even when the input is only composed of noise
Solution Approach 1:
The patent segments the estimation process into distinct steps: computing conditional expectations for individual signals, determining support regions, and reconstructing the final signal. This segmentation allows efficient processing while maintaining accuracy by treating signal presence detection and amplitude estimation as separate operations.
Solution Approach 2:
The patent replaces traditional mechanical linear filtering operations with probabilistic conditional expectation computations. This substitution enables the system to achieve both computational efficiency and accurate signal detection by using probability theory instead of conventional filtering mechanics.
Data Source
Figure 1
Figure 2~3
AI summary
This method for estimating at most two mixed signals from separate sources, of which the time-frequency representation shows an unknown non-zero proportion of zero components, by means of an array consisting of P>2 antennas, when the directional vectors U and V of the sources emitting said signals are additionally known or estimated, comprises the following steps: a) calculating the successive discrete Fourier transforms of the signal received by the antennas and sampled to obtain a time-frequency P-vector grid of the signal; each element of the grid being referred to as a box and containing a complex vector X forming a measurement; b) for each box, calculating the conditional expectation estimator of the signal, or of the signals, from the measurement X and an a priori probability density for the signals that is a Gaussian mixture.