Cognitive blind source separator

Patent metadata

Field	Value
Publication number	US-9749007-B1
Application number	US-201615073626-A
Country	US
Kind code	B1
Filing date	Mar 17, 2016
Priority date	Mar 19, 2015
Publication date	Aug 29, 2017
Grant date	Aug 29, 2017

Abstract

Described is a cognitive blind source separator (CBSS). The CBSS includes a delay embedding module that receives a mixture signal (the mixture signal being a time-series of data points from one or more mixtures of source signals) and time-lags the signal to generate a delay embedded mixture signal. The delay embedded mixture signal is then linearly mapped into a reservoir to create a high-dimensional state-space representation of the mixture signal. The state-space representations are then linearly mapped to one or more output nodes in an output layer to generate pre-filtered signals. The pre-filtered signals are passed through a bank of adaptable finite impulse response (FIR) filters to generate separate source signals that collectively formed the mixture signal.

First claim

What is claimed is:

1. A cognitive blind source separator (CBSS), the CBSS comprising:
one or more processors and a memory, the memory being a non-transitory computer-readable medium having executable instructions encoded thereon, such that upon execution of the instructions, the one or more processors perform operations of:
receiving a mixture signal, the mixture signal being a time-series of data points from one or more mixtures of source signals;
generating a delay embedded mixture signal based on the mixture signal;
linearly mapping the delay embedded mixture signal into a reservoir;
creating a high-dimensional state-space representation of the mixture signal by combining the delay embedded mixture signal with reservoir states;
identifying at least one of the separate source signals that collectively form the mixture signal based on the high-dimensional state-space representation of the mixture signal; and
linearly mapping the state-space representations to one or more output nodes in an output layer to generate ore-filtered signals, wherein each node in the output layer has a unique and adaptable set of output weights on its input connections, such that as one or more mixture signals are passed into the CBSS, the output weights are adapted so that output from a given node amplifies a subset of the source signals while suppressing others.

2. The cognitive blind source separator as set forth in claim 1 , wherein at least one of the separate source signals is identified by adaptively filtering the pre-filtered signals.

3. The cognitive blind source separator as set forth in claim 2 , wherein in adaptively filtering, the pre-filtered signals are passed through a bank of adaptable finite impulse response (FIR) filters.

4. The cognitive blind source separator as set forth in claim 3 , wherein each pre-filtered signal is fed into a unique filter in the bank of adaptable FIR filters, with each unique filter having an adaptable center frequency.

5. The cognitive blind source separator as set forth in claim 4 , wherein the output layer comprises a set of summing nodes, each summing node receiving a weighted output from each reservoir node as input, and summing these values together to produce its output.

6. The cognitive blind source separator as set forth in claim 5 , further comprising an operation of deriving an error signal, with the error signal being used to update the center frequencies of the filters in the bank of adaptable FIR filters and update the output weights in the output layer.

7. The cognitive blind source separator as set forth in claim 6 , wherein each output node in the output layer of the reservoir is associated with a unique FIR filter and the pre-filtered signals are passed through their respective FIR filter.

8. The cognitive blind source separator as set forth in claim 7 , further comprising an operation of extracting features from at least one of the separate source signals.

9. The cognitive blind source separator as set forth in claim 8 , wherein the reservoir is a recurrent neural network with a plurality of nodes.

10. The cognitive blind source separator as set forth in claim 1 , further comprising an operation of extracting features from at least one of the separate source signals.

11. The cognitive blind source separator as set forth in claim 1 , wherein the reservoir is a recurrent neural network with a plurality of nodes.

12. A cognitive blind source separator (CBSS), the CBSS comprising:
a delay embedder for receiving a mixture signal and generating a delay embedded mixture signal, the mixture signal being a time-series of data points from one or more mixtures of source signals;
a reservoir, the reservoir adapted to linearly map the delay embedded mixture signal into the reservoir to create a high-dimensional state-space representation of the mixture signal;
an output layer, the output layer adapted to generate pre-filtered signals by linearly mapping the state-space representations to one or more output nodes in the output layer, wherein each node in the output layer has a unique and adaptable set of output weights on its input connections, such that as one or more mixture signals are passed into the CBSS, the output weights are adapted so that output from a given node amplifies a subset of the source signals while suppressing others; and
a bank of adaptable finite impulse response (FIR) filters, the bank of FIR filters adapted to generate separate source signals that collectively form the mixture signal by processing the pre-filtered signals through the bank of adaptable FIR filters.

13. A computer program product for generating separate source signals from a mixture signal using a cognitive blind source separator (CBSS), the computer program product comprising:
a non-transitory computer-readable medium having executable instructions encoded thereon, such that upon execution of the instructions by one or more processors, the one or more processors perform operations of:
receiving a mixture signal, the mixture signal being a time-series of data points from one or more mixtures of source signals;
generating a delay embedded mixture signal based on the mixture signal;
linearly mapping the delay embedded mixture signal into a reservoir;
creating a high-dimensional state-space representation of the mixture signal by combining the delay embedded mixture signal with reservoir states;
identifying at least one of the separate source signals that collectively form the mixture signal based on the high-dimensional state-space representation of the mixture signal; and
linearly mapping the state-space representations to one or more output nodes in an output layer to generate ore-filtered signals, wherein each node in the output layer has a unique and adaptable set of output weights on its input connections, such that as one or more mixture signals are passed into the CBSS, the output weights are adapted so that output from a given node amplifies a subset of the source signals while suppressing others.

14. The computer program product as set forth in claim 13 , wherein at least one of the separate source signals is identified by adaptively filtering the pre-filtered signals.

15. The computer program product as set forth in claim 14 , wherein in adaptively filtering, the pre-filtered signals are passed through a bank of adaptable finite impulse response (FIR) filters.

16. The computer program product as set forth in claim 15 , wherein each pre-filtered signal is fed into a unique filter in the bank of adaptable FIR filters, with each unique filter having an adaptable center frequency.

17. A method for generating separate source signals from a mixture signal using a cognitive blind source separator (CBSS), the method comprising an act of:
causing one or more processors to execute instructions encoded on a non-transitory computer-readable medium, such that upon execution, the one or more processors perform operations of:
receiving a mixture signal, the mixture signal being a time-series of data points from one or more mixtures of source signals;
generating a delay embedded mixture signal based on the mixture signal;
linearly mapping the delay embedded mixture signal into a reservoir;
creating a high-dimensional state-space representation of the mixture signal by combining the delay embedded mixture signal with reservoir states;
identifying at least one of the separate source signals that collectively form the mixture signal based on the high-dimensional state-space representation of the mixture signal; and
linearly mapping the state-space rep

Assignees

• Hrl Lab Llc

Inventors

• Martin Charles E
• Rao Shankar R
• Petre Peter

Classifications

• H04B1/719 Primary

  Interference-related aspects · CPC title

• G06F2218/22

  Source localisation; Inverse modelling · CPC title

• G01S7/414 Primary

  Discriminating targets with respect to background clutter · CPC title

• G06N3/044

  Recurrent networks, e.g. Hopfield networks · CPC title

• G06F18/2134

  based on separation criteria, e.g. independent component analysis · CPC title