Processing device comprising a covariance matrix correction device

• US 6,721,693 B2
• Filed: 09/18/2002
• Issued: 04/13/2004
• Est. Priority Date: 09/20/2001
• Status: Active Grant

• Current Assignee: Raise Partner
• Original Assignee: Raise Partner
• Inventors: Oustry, François
• Primary Examiner(s): Hoff, Marc S.
• Assistant Examiner(s): Tsai, Carol
• Application Number:
  US10/245,376
• Publication Number:
  US 20030088386A1
• Time in Patent Office: 573 Days
• Field of Search: 702/27, 702/85, 702/185, 702/189, 702/190, 702/191, 702/192, 702/193, 702/194, 702/196, 342/80, 342/147, 342/148, 342/149, 342/152, 342/194, 342/195, 342/16, 342/90, 342/77, 342/417, 342/427, 342/373, 342/378, 706/22, 706/25
• US Class Current: 702/196
• CPC Class Codes: G06F 17/10 Complex mathematical operations function generation by table look-up G06F1/03; evaluation of elementary functions by calculation G06F7/544

Abstract

The correction device comprises a block for decomposition into total proper elements, according to the Kato spectral decomposition method, of the covariance matrix to be corrected. The block supplies to a correction block the total proper values and the total projectors of the covariance matrix, as well as first coefficient vectors for the combination of the source signals of the covariance matrix. The correlation block estimates the reconstructed variances of combinations of the source signals using the first coefficient vectors and second used-defined coefficient vectors. It looks for the corrected matrix closest to the matrix to be corrected, in relation to the variances and total projectors while is applying a quasi-Newton type method to the dual of a semi-defined lesser error square programme.

Claims

1. A processing device for a plurality of source digital signals, composed of measurement samples of the signals to be analysed, comprising a correction device for a covariance matrix, representative of the relations between the source signals, whereas the correction device comprises means for decomposing into total proper elements, according to the Kato spectral decomposition method, supplying the total proper values of the covariance matrix to be corrected and of the first coefficient vectors for the combination of the source signals, whereby the correction device comprises means of correction using the total proper values and the first coefficient vectors in order to determine the corrected matrix.

2. A device according to claim 1, wherein the correction device comprises user-defined means of determination of second coefficient vectors for the combination of the source signals.

3. A device according to claim 1, wherein the correction means comprises means for assessing combination variances of the source signals using the first and/or the second coefficient vectors.

4. A device according to claim 3, wherein the correction device comprises searching means for the corrected matrix closest to the matrix to be corrected, in relation to the variances supplied by the assessment means and to total projectors supplied by the means of decomposition into total proper elements.

5. A device according to claim 4, wherein the searching means apply a quasi-Newton type method on the dual of a semi-defined least error square programme.

6. A device according to claim 5, wherein the searching means apply a BFGS type method.

7. A device according to claim 3, wherein the assessment means of the variances comprise means for decomposing coefficient vectors into positive and negative portions.