Patent №
US 5,209,237
Granted
—
Owner
—
Lab
—
AI components
2
vision · planning
Assignment
None on record
Dataset
AIPD
2023_r1 edition
Application
07705493
To provide an accurate reproduction of a signal detected in a noisy environment, a primary signal sensor and a plurality of "signal-free" reference sensors are used. Information collected from the primary sensor and the reference sensors is compared to derive correlation data. A matrix, representing correlation data in the time domain or cross-spectral density information in the frequency domain, is constructed based on the computed correlation data. The matrix is transformed to a diagonal matrix using a singular value decomposition algorithm and a threshold value is selected with reference to one of the larger eigenvalues of the diagonalized matrix. Components of the diagonalized matrix are compared to the threshold and values greater than the threshold are used together with primary sensor correlation data to compute Wiener filter constants. The reference sensor outputs are modified by the filter constants and modified sensor outputs are subtracted from the primary sensor output. A noise-cancelled output signal is then synthesized. By selecting an appropriate threshold, the effect of redundant reference sensor contributions and the effect of small primary signal contributions to the reference sensors are eliminated.