Abstract: Echo cancellation for a two-way audio communication includes receiving, at an AEC system from microphone(s), an audio signal based on, at least in part, near-end signals and reproduced far-end signals. Loudspeaker(s) reproduced the far-end signals. The AEC system is operated, at least in part, with filter(s) so as to update estimates of coefficients of an acoustic channel from the loudspeaker(s) to the microphone(s). Control parameter(s) affecting an operation of the AEC system that is/are configurable and is/are set to value(s), from a range of values, is/are determined, based on estimating an accuracy of the estimates of the coefficients of the acoustic channel and a characteristic of the near-end signals. The AEC system controls the filter(s) with different values of the control parameter(s) at different times.

Abstract: An AEC system can implement different updating techniques for echo cancellation. A Block Maximum Likelihood updating technique updates filter coefficients typically at integer multiples of a block size. A Proportional-Incremental Maximum Likelihood updating technique applies proportional weighting to the updates of filter coefficients. A Multi-source Incremental Maximum Likelihood updating technique uses separate control parameters associated with each of multiple loudspeakers. Combinations of these may be used, and one or more of these can be selected and implemented in real-time.

Abstract: Echo cancellation for a two-way audio communication includes receiving, at an AEC system from microphone(s), an audio signal based on, at least in part, near-end signals and reproduced far-end signals. Loudspeaker(s) reproduced the far-end signals. The AEC system is operated, at least in part, with filter(s) so as to update estimates of coefficients of an acoustic channel from the loudspeaker(s) to the microphone(s). Control parameter(s) affecting an operation of the AEC system that is/are configurable and is/are set to value(s), from a range of values, is/are determined, based on estimating an accuracy of the estimates of the coefficients of the acoustic channel and a characteristic of the near-end signals. The AEC system controls the filter(s) with different values of the control parameter(s) at different times.

Abstract: Systems and methods for characterizing networks are disclosed. In several embodiments, a network analyzer applies a network analysis to a network that replaces components of the network in a model of the network with equivalent or bounding models. The network analyzer can then characterize the simplified model of the network and an assessment can be made concerning the accuracy of the characterization of the network.

Abstract: A method is provided for decompressing images captured by a compressive imaging system in conjunction with a compressive imaging application. The method comprises determining an estimation error et based on a difference between: (i) a measurement y captured by the compressive imaging system and (ii) a product of a sensing matrix H and a signal xt that has been estimated. The method comprises determining a gradient descent st+1 based on a sum of: (i) the signal xt that has been estimated and (ii) a product of a step size ?, a vector formulation of the sensing matrix HT and the estimation error et. The method comprises projecting the gradient descent st+1 comprising applying a compression function f and a decompression function g to estimate the signal xt+1.

Abstract: A method including, at a processor, obtaining an input to be classified, training a neural network, and classifying the input using the trained neural network. The training of the neural network being done in a feed-forward manner based on a plurality of sample inputs and sample classifications. Each of the sample classifications is associated with one of the plurality of sample inputs. The training includes, for each of a plurality of layers in the neural network, selecting values for a weight matrix, parameters for a plurality of activation unit functions, and a plurality of linear filter parameters. The classifying the input using the trained neural network is based on the weight matrix, the parameters for the plurality of activation unit functions, and the plurality of linear filter parameters for each of the plurality of layers.

Abstract: Systems and methods for characterizing networks are disclosed. In several embodiments, a network analyzer applies a network analysis to a network that replaces components of the network in a model of the network with equivalent or bounding models. The network analyzer can then characterize the simplified model of the network and an assessment can be made concerning the accuracy of the characterization of the network.

Abstract: Systems and methods for characterizing networks are disclosed. In several embodiments, a network analyzer applies a network analysis to a network that replaces components of the network in a model of the network with equivalent or bounding models. The network analyzer can then characterize the simplified model of the network and an assessment can be made concerning the accuracy of the characterization of the network.

Abstract: The present invention provides methods for universal lossy compression that provide performance at or near the rate-distortion limit and that are based on universal, implementable lossy source coding algorithms.

Abstract: The present invention provides methods for universal lossy compression that provide performance at or near the rate-distortion limit and that are based on universal, implementable lossy source coding algorithms.