A simple, histogram-based pattern classifier can be used to model somatosensory cortical receptive field changes in monkeys after digit amputation. If a cortical neuron in an adult monkey has a receptive field on an amputated digit, the neuron's receptive field initially becomes much larger after amputation, but is eventually restricted to a small area on the side of a neighboring digit. These observations can be explained by a classifier in which: (1) cortical neurons represent spatially ordered hypotheses that a stimulus is present in a given area of the hand; (2) the hypothesis made by each cortical neuron is unaffected by amputation or equivalent damage; (3) hypotheses are tested with a maximum likelihood, histogram-based classifier; (4) the classifier is trained and updated by recursive Hebbian learning; (5) imperfect exemplars are used to train the classifier; and (6) lateral inhibition exists between sensory sites. Implications of the model for prosthesis are suggested.>

A maximum likelihood (ML) classifier for discriminating between nonstationary Gaussian time series can be implemented by correlating the data spectrogram with templates that are constructed from ensemble average reference spectrograms. The time window used to synthesize the spectrograms must have a duration that is longer than the decorrelation time of the data in the neighborhood of the window. If the data time series exhibits significant nonstationarity within this decorrelation time, Karhunen-Loéve (K-L) basis functions should ideally be used to construct a generalized spectrogram, rather than using a standard spectrogram constructed with the usual sinusoidal basis functions. Utilization of a standard spectrogram imposes forced, pseudo-stationarity by approximating the autocovariance function of the data by the short-time autocorrelation function. This forced stationarity is routinely used to obtain linear prediction coefficients (LPC). When signal to interference ratio (SNR) is large, the templates that are used to classify a data spectrogram are sensitive to differences in the locations of nulls or zeroes in the expected signal spectrograms from different data classes. This null sensitivity seems to imply that peak-oriented models of random processes, e.g. , the all pole representation that is associated with LPC, are suboptimum for ML classification under high SNR conditions. Compensation for time warping is especially necessary if window durations are data dependent. Spectrogram implementation of the ML classifier yields a new similarity index for time warp compensation.

The Berry-Esseén theorem has been used to show that narrow-band filtering tends to make non-Gaussian noise more Gaussian. In this correspondence, the same theorem is invoked to demonstrate that wide-band filtering can also make noise more Gaussian, provided the filter impulse response has a large time-bandwidth product.

Methods for narrow-band clutter suppression are examined to determine their applicability to wide-band signal analysis. Algorithms amenable to wide-band theory are established, and a simple example is given.