Method and apparatus for situationally adaptive processing in echo-location systems operating in non-Gaussian environments
AssigneeMultispec Corporation
InventorsVasilis Z. Marmarelis, Chrysostomos L. Nikias, Dae Cheol Shin
Life of this patent
Estimating the term…
Abstract
A signal processing method for use in echo-location systems (radar, sonar, laser radar) to obtain accurate estimates of the target reflectivity sequence in a random (non-Gaussian) noise environment, including ambient noise, reverberation, and clutter. The method derives the statistical characteristics of the random background environments and then constructs and maximizes the corresponding approximate likelihood function using iterative methods. The approximate maximum likelihood estimates are generated on the basis of an approximation of the ideal likelihood function which is maximized using computationally efficient algorithms.
Claims16
In an echo-location system, a method for adaptively estimating the target reflectivity sequence from received data that has been corrupted by non-Gaussian disturbances, said method comprising:
- estimating the statistics of the prevailing non-Gaussian disturbance in the received data;
- generating an approximate likelihood function from the estimated statistics of the prevailing non-Gaussian disturbances;
- and
- maximizing the likelihood function by applying an iterative gradient-descent algorithm to obtain the target reflectivity sequence.
The method of claim 1 or claim 3 wherein said estimating step comprises fitting the data to broad classes of parametric distributions to obtain a probability density function of the disturbances.
In an echo-location system, a method for adaptively estimating the target reflectivity sequence from received data that has been corrupted by non-Gaussian disturbances, said method comprising:
- estimating the statistics of the prevailing non-Gaussian disturbances in the received data;
- generating an approximate likelihood function from the estimated statistics of the prevailing non-Gaussian disturbances by obtaining a log-likelihood function;
- and
- maximizing the likelihood function to obtain the target reflectivity sequence.
In a data reception system, a method for adaptively estimating the transmitted data from received data that has been corrupted by non-Gaussian disturbances, said method comprising:
- estimating the statistics of the prevailing non-Gaussian disturbances in the received data;
- generating an approximate likelihood function from the estimated statistics of the prevailing non-Gaussian disturbances;
- and
- maximizing the likelihood function by applying an iterative gradient-descent algorithm to obtain the transmitted data.
The method of claim 6 or claim 8 wherein said estimating step comprises fitting the data to broad classes of parametric distributions to obtain a probability density function of the disturbances.
In a data reception system, a method for adaptively estimating the transmitted data from received data that has been corrupted by non-Gaussian disturbances, said method comprising:
- estimating the statistics of the prevailing non-Gaussian disturbances in the received data;
- generating an approximate likelihood function from the estimated statistics of the prevailing non-Gaussian disturbances by obtaining a log-likelihood function;
- and
- maximizing the likelihood functions to obtain the transmitted data.
In an echo location system, a method for adaptively estimating target-channel kernels h(k) from given transmitted data x(k) and given received data y(k) in the presence of severe noise, clutter and reverberation .epsilon.(k), said parameters having the relationship y(k)=h(k)*x(k)+.epsilon.(k), wherein * denotes convolution, said method comprising;
- estimating the statistics of .epsilon.(k) to obtain a probability density function p(.epsilon.);
- constructing an approximate likelihood function based on p(.epsilon.) for N independent samples according to the relationship:
- ##EQU9## where:
- ##EQU10## and y(n) and x(n) are the n-th samples of the respective received and transmitted signals, and h(m) is the sought unknown discrete kernel values;
- and
- maximizing the likelihood function to obtain the estimated h(m).
Dependencies are inferred from the claim text.