System and methods for autonomous signal modulation format identification

Patent metadata

Field	Value
Publication number	US-9768874-B1
Application number	US-201414587306-A
Country	US
Kind code	B1
Filing date	Dec 31, 2014
Priority date	Mar 20, 2013
Publication date	Sep 19, 2017
Grant date	Sep 19, 2017

Abstract

Systems and methods for autonomous signal modulation format identification are disclosed. In an example embodiment of the disclosed technology, a method includes mapping an input signal to Stokes space to generate a representation of the input signal in three-dimensional space. The method may further include determining the dimension of the representation and, based on the dimension, selecting a subset of modulation from a plurality of mutually exclusive subsets of modulation formats. Further, the method may include defining a cost function for identifying the modulation format from the selected subset and evaluating the cost function to identify the modulation format.

First claim

What is claimed is:

1. A method for autonomously determining a modulation format of an optical signal, the method comprising:
responsive to receiving the optical signal from a transmitter, converting the optical signal into an input signal;
mapping the input signal to Stokes space to generate a representation of the input signal in three-dimensional space, the representation taken to follow a Gaussian mixture model;
determining the dimension of the representation of the input signal in three-dimensional space;
based on the dimension of the representation of the input signal in three-dimensional space having two dimensions in Stokes space:
defining at least one cost function for identifying the modulation format of the input signal in from among OOK, BPSK, QPSK, M-PAM, and M-PSK, the cost function based at least in part on concentration parameters of mixing probabilities of the Gaussian mixture; and
evaluating the at least one cost function to identify the modulation format of the input signal from among OOK, BPSK, QPSK, M-PAM, and M-PSK;

based on the dimension of the representation of the input signal in three-dimensional space having three dimensions in Stokes space, identifying the modulation format of the input signal to be M-QAM for M≠{2;4}.

2. The method of claim 1 , wherein the representation of the input signal comprises a particular number of clusters in three-dimensional space.

3. The method of claim 2 , wherein each cluster in three-dimensional space comprises:
a weight;
a mean position; and
a particular number of Stokes space symbols.

4. The method of claim 3 , wherein each Stokes space symbol is a three-dimensional point in Stokes space and has three-dimensional coordinates.

5. The method of claim 4 further comprising:
modeling a probabilistic weight distribution of at least each of the particular number of clusters in three-dimensional space as a joint Dirichlet distribution to provide a joint Dirichlet distribution model of the probabilistic weight distribution; and
modeling a probabilistic mean position distribution of at least each of the particular number of clusters in three-dimensional space as a normal distribution to provide a normal distribution model of the probabilistic mean position distribution.

6. The method of claim 5 further comprising:
applying a Variational Bayesian method to the representation of the input signal in three-dimensional space to estimate:
a plurality of concentration parameters of the joint Dirichlet distribution model of the probabilistic weight distribution; and
a plurality of hypermean parameters of the first normal distribution model of the probabilistic mean position distribution.

7. The method of claim 6 , wherein each concentration parameter of the joint Dirichlet distribution model represents a weight of each of the particular number of clusters in three-dimensional space, and wherein each hypermean parameter of the first normal distribution model represents a mean position of each of the particular number of clusters in three-dimensional space.

8. The method of claim 6 further comprising:
generating a two-dimensional plane in Stokes space; and
identifying a normal to the two-dimensional plane, wherein the normal goes through the origin of the two-dimensional plane.

9. The method of claim 8 , wherein determining the dimension of the representation of the input signal comprises:
defining a cost function based on:
the normal to the two-dimensional plane; and
the plurality of hypermean parameters of the normal distribution model; and

evaluating the cost function to generate a value that serves as a proxy for the dimension of the representation of the input signal in three-dimensional space.

10. The method of claim 9 , wherein the joint Dirichlet distribution model of the probabilistic weight distribution is a first joint Dirichlet distribution model of the probabilistic weight distribution and the normal distribution model of the probabilistic mean position distribution is a first normal distribution model of the probabilistic mean position distribution, and wherein defining the at least one cost function for identifying the modulation format from among OOK, BPSK, QPSK, M-PAM, and M-PSK comprises:
projecting the representation of the input signal in three-dimensional space onto the two-dimensional plane to generate a two-dimensional projection of the representation of the input signal in three-dimensional space, wherein the two-dimensional projection comprises a particular number of clusters in two-dimensional space, and wherein each cluster in two-dimensional space comprises:
a weight;
a particular number of two-dimensional projected Stokes space symbols, wherein the particular number of two-dimensional projected Stokes space symbols is a proxy for the weight of the particular cluster in two-dimensional space, and wherein each two-dimensional projected Stokes space symbol is a two-dimensional point in the two-dimensional plane; and
a mean position;

modeling a probabilistic weight distribution of at least each of the particular number of clusters in two-dimensional space as a joint Dirichlet distribution to provide a second joint Dirichlet distribution model of the probabilistic weight distribution;
modeling a probabilistic mean position distribution of at least each of the particular number of clusters in two-dimensional space as a normal distribution to provide a second normal distribution model of the probabilistic mean position distribution;
applying a Variational Bayesian method to the two-dimensional projection to estimate:
a plurality of concentration parameters of the second joint Dirichlet distribution model of the probabilistic weight distribution, wherein each concentration parameter of the second joint Dirichlet distribution model of the probabilistic weight distribution is a proxy for the weight of each cluster of the particular number of clusters in two-dimensional space; and
a plurality of hypermean parameters of the second normal distribution model of the probabilistic mean position distribution, wherein each hypermean parameter of the second normal distribution model of the probabilistic mean position distribution is a proxy for the mean position of each cluster of the particular number of clusters in two-dimensional space;

defining higher-order-statistics relating to the two-dimensional projection, wherein higher-order-statistics are based on the plurality of concentration parameters of the second joint Dirichlet distribution model of the probabilistic weight distribution and the plurality of hypermean parameters of the second normal distribution model of the probabilistic mean position distribution;
evaluating the higher-order statistics to generate a plurality of higher-order cumulant values indicative of a particular modulation format signature within the first subset; and
responsive to generating the plurality of higher-order cumulant values, employing a decision tree to determine the modulation format of the input signal from among OOK, BPSK, QPSK, M-PAM, and M-PSK.

11. A method for autonomously determining a modulation format of an optical signal, the method comprising:
responsive to receiving the optical signal from a transmitter, converting the optical signal into an input signal;
mapping the input signal to Stokes space to generate a representation of the input signal in three-dimensional space, the representation of the input signal having a dimension and comprising a particular number of clusters in three-dimensional space, wherein each cluster in three-dimensional space comprises:
a weight;
a mean position; and
a particular number of Stokes space symbols, wherein

Assignees

• Georgia Tech Res Inst

Inventors

• Ralph Stephen E
• Isautier Pierre Paul Roger

Classifications

• H04B10/0795 Primary

  Performance monitoring; Measurement of transmission parameters · CPC title

• H04B10/516 Primary

  Details of coding or modulation · CPC title

• H04B10/616 Primary

  Details of the electronic signal processing in coherent optical receivers · CPC title

• H04B10/695

  Arrangements for optimizing the decision element in the receiver, e.g. by using automatic threshold control · CPC title

• H04L27/0012

  arrangements for identifying the type of modulation · CPC title