Abstract
An embodiment generally relates to a method of signal discrimination. The method includes receiving plurality of signal mixtures, where each signal has a property whose value is based on a principal eigenvalue of a unique separating operator. The method also includes creating a separating operator using a target signal approximation and applying a separating operator for a target signal to the plurality of signal mixtures. The method further includes obtaining the target signal based on an operation of optimization of the separating operator for maximum property value amplitude on the plurality of signal mixtures.
Claims

A method of signal separation and discrimination, the method comprising:
receiving a plurality of signal mixtures at a processor, each signal having a property whose value is based on a principal eigenvalue;
creating a signal separating operator, F_{s }using a target signal approximation s_{a};
applying the signal separating operator to the plurality of signal mixtures; and
obtaining the target signal s from the plurality of signal mixtures based on the application of the signal separating operator on the plurality of signal mixtures.

The method of claim 1, wherein the applying the signal separating operator, further comprises:
step a. performing a singular value decomposition of the plurality of signal mixtures;
step b. storing a plurality of signal subspace vectors based on the singular decomposition of the plurality of signal mixtures as a matrix V; and
step c. obtaining the principal eigenvector of the signal separating operator restricted to a signal subspace spanned by a plurality of signal subspace vectors determined from the plurality of signals.

The method of claim 2, further comprising:
step d. updating the signal vector, s, with the principal eigenvector of the basis matrix, V^{(U)}, wherein V^{(U)}=VU_{s }and U_{s }is based on a principal component decomposition, U_{s}λ(s)U_{s}^{T }of V^{T}F_{s}V such that V^{(U)λ V}^{(U)T}=VU_{s}λU_{s}^{T }V^{T}=V V^{T }F_{s }V V^{T}, is the principal component decomposition of the restriction of the signal separating operator to the signal space of the plurality of signal mixtures where V V^{T }is a signal space projection operator.

The method of claim 3, further comprising:
step e. updating the matrix V with rotated signal subspace basis vectors, V^{(U)}=VU_{s }and updating the property operator with the property operator corresponding to the principal eigenvector of V^{(U)}; and
step f. repeating steps de until an equality condition is achieved.
 The method of claim 4, wherein the equality condition is a rotation matrix, U_{s}, approximately equal to an identity matrix.

The method of claim 4, further comprising:
step g. creating a deflated signal subspace by removing, s_{max}, from the rotated subspace basis vectors; and
step h. repeating steps dg until all signals have been sequentially removed from the deflated signal subspace basis vectors and placed in a signal matrix, S, arranged by signal property amplitude.

The method according to claim 3, further comprising:
step e. updating the matrix V with rotated signal subspace basis vectors, V^{(U)}=VU_{s }and updating the property operator with the property operator corresponding to the principal eigenvector of V^{(U)};
step f. repeating steps de N times, wherein N is determined by the number of removals of the partial signal estimates in step i;
step g. removing a partial signal estimate of s_{max}, from the rotated subspace basis vectors;
step h. repeating steps dg until all partial signal estimates of s_{max }have been sequentially removed from the deflated signal subspace basis vectors and placed in a signal matrix, V^{(est)}, arranged by signal property amplitude; and
step i. updating the signal basis spanning vectors, V, with the partial signal estimate vectors, V^{(est)}.

The method according to claim 7, further comprising:
step j. repeating steps di until all the partial signal estimates included in the matrix, V^{(est)}, are approximately equal to the prior partial signal estimates of, V^{(est)}, obtained before repeating steps di; and
step k. setting signal estimates S equal to the partial signal estimates of the matrix, V^{(est)}, arranged by signal property amplitude.
 The method of claim 1, wherein the plurality of signal mixtures is associated with electromagnetic signals.
 The method of claim 1, wherein the plurality of signal mixtures is associated with nonelectromagnetic signals.
 The method of claim 1, wherein the plurality of signal mixtures is associated with brain functions.
 The method of claim 1, wherein the plurality of signal mixtures is timedependent.
 The method of claim 1, wherein the plurality of signal mixtures is timeindependent.
 The method of claim 1, wherein the method is applied to one of the following types of signal discrimination problems, including blind separation of signals, encryption, artifact separation of signals, separation of burst signals, or noise filtering.

A computer storage device embedded with one or more computer programs, said one or more computer programs implementing a method of signal discrimination, said one or more computer programs comprising a set of instructions for:
receiving a plurality of signal mixtures, each signal having a property whose value is based on a principal eigenvalue;
creating a signal separating operator using a target signal approximation;
applying the signal separating operator to the plurality of signal mixtures; and
obtaining a target signal from the plurality of signal mixtures based on the application of the signal separating operator on the plurality of signal mixtures.
 The computer storage device of claim 15, wherein the embedded one or more computer programs is executed to solve one of the following types of signal discrimination problems, including blind separation of signals, encryption, artifact separation of signals, separation of burst signals, or noise filtering.

The computer storage device according to claim 15, wherein said set of instructions further comprises:
performing a singular value decomposition of the plurality of signal mixtures;
storing a plurality of signal subspace vectors based on the singular decomposition of the plurality of signal mixtures as a matrix V; and
obtaining the principal eigenvector of the signal separating operator restricted to a signal subspace spanned by a plurality of signal subspace vectors determined from the plurality of signals.

The computer storage device according to claim 17, wherein said set of instructions further comprises:
updating the signal vector, s, with the principal eigenvector of the basis matrix, V^{(U)}, wherein V^{(U)}=VU_{s }and U_{s }is based on a principal component decomposition, U_{s}λ(s)U_{s}^{T }of V^{T}F_{S}V such that V^{(U)}λ V^{(U)T}=VU_{sλU}_{s}^{T }V^{T}=V V^{T }F_{s }V V^{T}, is the principal component decomposition of the restriction of the signal separating operator to the signal space of the plurality of signal mixtures where V V^{T }is a signal space projection operator.

The computer storage device according to claim 18, wherein said set of instructions further comprises:
updating the matrix V with rotated signal subspace basis vectors, V^{(U)}=VU_{s }and updating property operator F_{s }with the property operator corresponding to the principal eigenvector of V^{(U)}; and
repeating the updating steps until an equality condition is achieved.
 The computer storage device according to claim 19, wherein the equality condition is a rotation matrix, U_{s}, and is approximately equal to an identity matrix.

The computer storage device according to claim 19, wherein said set of instructions further comprises:
creating a deflated signal subspace by removing s_{max}, from the rotated subspace basis vectors; and
repeating the updating, repeating, and creating steps until all signals have been sequentially removed from the deflated signal subspace basis vectors and placed in a signal matrix, S, arranged by signal property amplitude.

The computer storage device according to claim 18, wherein said set of instructions further comprises:
updating the matrix V with rotated signal subspace basis vectors, V^{(U)}=VU_{s }and updating property operator F_{s }with the property operator corresponding to the principal eigenvector of V^{(U)};
repeating the updating steps N times, wherein N is determined by the number of removals of the plurality of signals in the following updating step;
removing a partial estimate of s_{max}, from the rotated subspace basis vectors;
repeating the updating, repeating, and removing steps until all partial signal estimates of s_{max }have been removed from the deflated signal subspace basis vectors and placed in a signal matrix, V^{(est)}, arranged by signal property amplitude; and
updating the signal basis spanning vectors, V, with the partial signal estimate vectors, V^{(est)}.

The computer storage device according to claim 22, wherein said set of instructions further comprises:
repeating the updating, repeating, removing, and updating steps until all the partial signal estimates included in the matrix, V^{(est)}, are approximately equal to the prior partial signal estimates of V^{(est)}, obtained before repeating the updating, repeating, removing, and updating steps; and
setting the partial signal estimates equal to the partial signal estimates of the matrix, V^{(est)}, arranged by signal property amplitude.

A method of signal separation and discrimination, the method comprising:
receiving a plurality of signal mixtures at a processor, each signal having a property whose value is based on a principal eigenvalue;
creating a signal separating operator using a target signal property;
applying the signal separating operator to the plurality of signal mixtures; and
obtaining a target signal from the plurality of signal mixtures based on the application of the signal separating operator on the plurality of signal mixtures.

The method of claim 24, further comprising:
recursively optimizing a property operator until property operator is equal to the signal separating operator and configured to determine the target signal, s, wherein, the property amplitude of the target signal s is greater than all other signals in the signal space spanned by the matrix, V.
 The method of claim 24, wherein the method is applied to one of the following types of signal discrimination problems, including blind separation of signals, encryption, artifact separation of signals, separation of burst signals, or noise filtering.

A computer storage device embedded with one or more computer programs, said one or more computer programs implementing a method of signal discrimination, said one or more computer programs comprising a set of instructions for:
receiving a plurality of signal mixtures, each signal having a property whose value is based on a principal eigenvalue;
creating a signal separating operator using a target signal property;
applying the signal separating operator to the plurality of signal mixtures; and
obtaining a target signal from the plurality of signal mixtures based on the application of the signal separating operator on the plurality of signal mixtures.

The computer storage device according to claim 27, wherein said set of instructions further comprises:
recursively optimizing a signal property operator, F_{s}, until F_{s }is equal to the signal separating operator and configured to determine the target signal, s, wherein, the property amplitude of the target signal is greater than all other signals in the signal space spanned by the matrix, V.
 The computer storage device of claim 27, wherein the embedded one or more computer programs is executed to solve one of the following types of signal discrimination problems, including blind separation of signals, encryption, artifact separation of signals, separation of burst signals, or noise filtering.
Owners (US)

Signalspace Inc
(Mar 13 2006)
Explore more patents:
Applicants

Signalspace Inc
Explore more patents:
Inventors

Moran John E
Explore more patents:
CPC Classifications

G10L21/0272
Explore more patents:

G06K9/624
Explore more patents:
IPC Classifications

H04L27/06
Explore more patents:
US Classifications

375/340
Explore more patents:
Document Preview
 Publication: Apr 27, 2010

Application:
Mar 13, 2006
US 37332506 A

Priority:
Mar 13, 2006
US 37332506 A

Priority:
May 19, 2005
US 68240405 P