Abstract: A method for localizing signals of interest includes initializing characteristics of the signals. Signals are acquired from a sensor array having at least three acoustic sensors. After digitization and conditioning, the signals associated with each sensor are validated by comparison with initialized characteristics. The signals are correlated across sensor groups to obtain time differences of arrival (TDOA). These TDOA are validated and associated with other TDOA from different times. TDOA from different sensor pairs are associated when they share a common sensor. A hyperbola of possible locations is created for each validated TDOA. Summation of the hyperbolas gives an intensity function. The location is identified as the most intense point in the intensity function. The source can be tracked across time as a computer output.

Abstract: A method of selecting an optimal window length in a digital processing system includes receiving a digital signal and analyzing the signal with a group of Hanning windows having different sizes. The windows being arranged so that they can be scaled to be comparable. The digital signal is windowed with each selected window and a transform is computed. The transformed signal is scaled and corrected. A metric is computed from the resulting signal for each window. The metrics are compared and the window size is selected based on agreement with a user defined metric. The specific window function, shift and scaling are such that the resulting analysis is mathematically equivalent across different window sizes.

Abstract: A computer implemented method for determining frequency and slope of a linear frequency modulated signal. Windowed signal data is collected and used for formulating a search surface function which can be calculated recursively. The function is divided into parts. Part values are calculated at an underspecified number of points corresponding to the number of parts. The search surface function part values are interpolated at values between frequency values. An interpolation geometry is chosen and then the function part values are interpolated at values between frequency values and slopes. Function parts are added together giving a restored search surface function. The frequency and slope values can be searched to find the values that maximize the search surface function. Other embodiments result in faster execution.

Abstract: A method for classifying data includes selecting an elemental size and features for the data that are representative of possible subclasses. Resolution widths are selected in conjunction with these features. Models associated with symbols are developed from these resolution widths and features. Data is compared with these models to give a likelihood that the model applies. The best model is determined and a signal is provided related to the symbol associated with the best model.

Abstract: A method is provided for calculating a class-specific iterated subspace for a classification system utilized in a computing system. Training data in the specific class for the class-specific iterated subspace is collected. A linear orthogonal transform is applied transforming the data into at least one bin. Magnitude squared bins are calculated and used as columns of a matrix. Orthonormal vectors of this matrix are selected and a J function is calculated. The J function and orthonormal starting vectors are used to obtain the class-specific iterated subspace for each class. The method further applies these class-specific iterated subspaces in a classification system for determining the most likely class of a data signal of interest.

Abstract: A method for segmenting a signal into segments having similar spectral characteristics is provided. Initially the method generates a table of previous values from older signal values that contains a scoring value for the best segmentation of previous values and a segment length of the last previously identified segment. The method then receives a new sample of the signal and computes a new spectral characteristic function for the signal based on the received sample. A new scoring function is computed from the spectral characteristic function. Segments of the signal are recursively identified based on the newly computed scoring function and the table of previous values. The spectral characteristic function can be a selected one of an autocorrelation function and a discrete Fourier transform. An example is provided for segmenting a speech signal.

Abstract: Accordingly there is provided a class specific classifier for classifying data received from a data source. The classifier has a feature transformation section associated with each class of data which receives the data and provides a feature set for the associated data class. Each feature transformation section is joined to a pattern matching processor which receives the associated data class feature set. The pattern matching processors calculate likelihood functions for the associated data class. One normalization processor is joined in parallel with each pattern matching processor for calculating an inverse likelihood function from the data, the associated class feature set and a common data class set. The common data class set can be either calculated in a common data class calculator or incorporated in the normalization calculation. The inverse likelihood function is then multiplied with the likelihood function for each associated data class.

Abstract: A method for segmenting a signal into segments having similar spectral characteristics is provided. Initially the method generates a table of previous values from older signal values that contains a scoring value for the best segmentation of previous values and a segment length of the last previously identified segment. The method then receives a new sample of the signal and computes a new spectral characteristic function for the signal based on the received sample. A new scoring function is computed from the spectral characteristic function. Segments of the signal are recursively identified based on the newly computed scoring function and the table of previous values. The spectral characteristic function can be a selected one of an autocorrelation function and a discrete Fourier transform. An example is provided for segmenting a speech signal.

Abstract: A system and method for training a class-specific hidden Markov model (HMM) is used for modeling physical phenomena, such as speech, characterized by a finite number of states. The method receives training data and estimates parameters of the class-specific HMM from the training data using a modified Baum-Welch algorithm, which uses likelihood ratios with respect to a common state (e.g., noise) and based on sufficient statistics for each state. The parameters are stored for use in processing signals representing the physical phenomena, for example, in speech processing applications. The modified Baum-Welch algorithm is an iterative algorithm including class-specific forward and backward procedures and HMM reestimation formulas.

Abstract: An improved sonar signal processor and display combining the use of both coherent and incoherent signal processors. In addition to a conventionally used matched filter detection processor, an incoherent signal processor comprising a cross-range energy filter and a down-range energy filter is used. The cross-range energy filter detects objects characterized by a narrow bearing response; whereas the down-range energy filter detects objects characterized by a narrow range response. The detection events resulting from the incoherent signal processor are displayed in a subdued color to prevent distraction from the primary display events and to reduce the false alarm rate by allowing the sonar operator to view events in the context of natural boundaries.

Abstract: This invention reduces the ripple component of current in the alternating current to, or direct current from, a common power source providing direct current power to a plurality of switched dc/dc power supplies by controlling the time of occurrence of switching of each of the supplies. Each dc/dc power supply is controlled by a switch pulse from a switching circuit which determines the occurrence and duration of time during which each power supply is switched "on". The time of occurence of the switch pulse of each switching circuit is determined by the application of an individual synchronizing pulse to each switching circuit. The combined effect of the ripple components of currents provided to the switched power supplies is minimized by summing these currents with appropriately phased synchronizing pulses applied to the switching circuits.