Abstract: Various methods and systems are provided for adaptive background estimation. In one embodiment, among others, a signal processing device includes at least one processing unit and a memory storing a background estimation system executable in the at least one processing unit. The background estimation system, when executed, causes the signal processing device to separate foreground objects from at least one image frame in a sequence of image frames based at least in part upon a correntropy cost function. In another embodiment, a method includes obtaining an image frame from a sequence of image frames; performing correntropy filtering on the image frame to separate foreground objects from the image frame; and providing the foreground objects for rendering. The correntropy filtering may adapt to compensate for diffusion effects in the image frame.

Abstract: Various methods and systems are provided for related to adaptive systems using correntropy. In one embodiment, a signal processing device includes a processing unit and a memory storing an adaptive system executable in the at least one processing unit. The adaptive system includes modules that, when executed by the processing unit, cause the signal processing device to adaptively filter a desired signal using a correntropy cost function. In another embodiment, a method includes adjusting a coefficient of an adaptive filter based at least in part on a correntropy cost function signal, providing an adaptive filter output signal based at least in part on the adjusted coefficient and a reference signal, and determining an error signal based at least in part on a received signal and the adaptive filter output signal.

Abstract: Various methods and systems are provided for classification using correntropy. In one embodiment, a classifying device includes a processing unit and memory storing instructions in modules that when executed by the processing unit cause the classifying device to adaptively classify a data value using a correntropy loss function. In another embodiment, a method includes adjusting a weight of a classifier based at least in part on a change in a correntropy loss function signal and classifying a data value using the classifier. In another embodiment, a method includes classifying a data value by predicting a label for the data value using a discriminant function, determining a correntopy statistical similarity between the predicted label and an actual label based at least in part on a correntropy loss function, and minimizing an expected risk associated with the predicted label based at least in part on a correntropy statistical similarity.

Abstract: Various methods and systems are provided for adaptive background estimation. In one embodiment, among others, a signal processing device includes at least one processing unit and a memory storing a background estimation system executable in the at least one processing unit. The background estimation system, when executed, causes the signal processing device to separate foreground objects from at least one image frame in a sequence of image frames based at least in part upon a correntropy cost function. In another embodiment, a method includes obtaining an image frame from a sequence of image frames; performing correntropy filtering on the image frame to separate foreground objects from the image frame; and providing the foreground objects for rendering. The correntropy filtering may adapt to compensate for diffusion effects in the image frame.

Abstract: Methods, systems, and computer-readable and executable instructions are provided for classifying an electrocardiogram (ECG) signal. Classifying an ECG signal can include analyzing the ECG signal using a stream of pulses generated by a sampler, extracting cardiac pulse features from a timing of the stream of pulses, and classifying the ECG signal based on the extracted cardiac pulse feature.

Abstract: Methods, systems, and computer-readable and executable instructions are provided for classifying an electrocardiogram (ECG) signal. Classifying an ECG signal can include analyzing the ECG signal using a stream of pulses generated by a sampler, extracting cardiac pulse features from a timing of the stream of pulses, and classifying the ECG signal based on the extracted cardiac pulse feature.

Abstract: Various methods and systems are provided for related to adaptive systems using correntropy. In one embodiment, a signal processing device includes a processing unit and a memory storing an adaptive system executable in the at least one processing unit. The adaptive system includes modules that, when executed by the processing unit, cause the signal processing device to adaptively filter a desired signal using a correntropy cost function. In another embodiment, a method includes adjusting a coefficient of an adaptive filter based at least in part on a correntropy cost function signal, providing an adaptive filter output signal based at least in part on the adjusted coefficient and a reference signal, and determining an error signal based at least in part on a received signal and the adaptive filter output signal.

Abstract: Various methods and systems are provided for classification using correntropy. In one embodiment, a classifying device includes a processing unit and memory storing instructions in modules that when executed by the processing unit cause the classifying device to adaptively classify a data value using a correntropy loss function. In another embodiment, a method includes adjusting a weight of a classifier based at least in part on a change in a correntropy loss function signal and classifying a data value using the classifier. In another embodiment, a method includes classifying a data value by predicting a label for the data value using a discriminant function, determining a correntopy statistical similarity between the predicted label and an actual label based at least in part on a correntropy loss function, and minimizing an expected risk associated with the predicted label based at least in part on a correntropy statistical similarity.

Abstract: A method of building a model for a physical plant in the presence of noise can include initializing the model of the physical plant, wherein the model is characterized by a parameter vector, estimating an output of the model, and computing a composite cost comprising a weighted average of an error between the estimated output from the model and an actual output of the physical plant, and a derivative of the error. The method further can include determining a step size and a model update direction. The model of the physical plant can be updated. The updating step can be dependent upon the step size. Another embodiment can include the steps of determining a Kalman gain and determining an error vector comprised of two entries weighted by a scalar parameter.

Abstract: An iterative method of equalizing an input signal received over a digital communication channel can include (a) using a kernel density estimate where different values of a kernel size are indicative of either a blind or a decision-directed equalization mode, (b) processing a received signal using a blind equalization mode, and (c) evaluating, on a block or sample basis, an error measure based on a distance among a distribution of an equalizer output and a constellation. The method also can include (d) updating the kernel size based upon the error measure thereby facilitating automatic switching between the blind and decision-directed equalization modes, where the kernel size is initially set to a value indicative of the blind equalization mode. The method additionally can include (e) selectively applying blind equalization or decision-directed equalization to the input signal according to the updated kernel size for subsequent iterations of steps (c)-(e).

Abstract: An iterative method of equalizing an input signal received over a digital communication channel can include (a) using a kernel density estimate where different values of a kernel size are indicative of either a blind or a decision-directed equalization mode, (b) processing a received signal using a blind equalization mode, and (c) evaluating, on a block or sample basis, an error measure based on a distance among a distribution of an equalizer output and a constellation. The method also can include (d) updating the kernel size based upon the error measure thereby facilitating automatic switching between the blind and decision-directed equalization modes, where the kernel size is initially set to a value indicative of the blind equalization mode. The method additionally can include (e) selectively applying blind equalization or decision-directed equalization to the input signal according to the updated kernel size for subsequent iterations of steps (c)-(e).