Abstract: A spatial image processor neural network for processing image data to discriminate between first and second spatial configurations of component objects includes a photo transducer input array for converting an input image to pixel data and sending the data to a localized gain network (LGN) module, a parallel memory processor and neuron array for receiving the pixel data and processing the pixel data into component recognition vectors and chaotic oscillators for receiving the recognition vectors and sending feedback data to the LGN module as attention activations. The network further includes a temporal spatial retina for receiving both the pixel data and temporal feedback activations and generating temporal spatial vectors, which are processed by a temporal parallel processor into temporal component recognition vectors. A spatial recognition vector array receives the temporal component recognition vectors and forms an object representation of the first configuration of component objects.

Abstract: A color sensor for generating color information defining colors of an image includes an input section, a color processing section, a color comparison section, a color boundary processing section and a memory processing section. The input section includes an array of transducer pairs, each pair defining one of a plurality of pixels. Each transducer pair generates two peak outputs, one for the selected color of each transducer of the pair. A plurality of pixel processors in the color processing section each receives the outputs from one of the transducer pairs. The color processing section generates a color feature vector representative of the brightness of the light incident on the pixels and a color value corresponding to the ratio of outputs from the transducers comprising the transducer pair associated with the pixels. The color boundary processing section generates a plurality of color boundary feature vectors, each representing the difference between the color value for a pixel and its neighboring pixels.

Abstract: A static memory processor for pattern recognition and an input data dimensionality reduction is provided having a multi-layer harmonic neural network and a classifier network. The multi-layer harmonic neural network receives a fused feature vector of the pattern to be recognized from a neural sensor and generates output vectors which aid in discrimination between similar patterns. The fused feature vector and each output vector are separately provided to corresponding positional king of the mountain (PKOM) circuits within the classifier network. Each PKOM circuit generates a positional output vector with only the element corresponding to the element of the fused feature vector or output vector having the highest contribution in its respective vector having a value corresponding to one. The positional output vectors are active in a multidimensional memory space and are read by a recognition vector array which generates class likelihood outputs determined by the occupied memory space.

Abstract: A multimode invariant processor is provided to simultaneously classify one or more patterns in multidimensional or in two dimensional “real world” images. The classification is invariant to a translation, a change in scale size and a rotation of a whole or partially hidden photonic image. The multimode invariant image processor comprises a retina portion, a nonlinear processing portion, a convergence processing portion and a classifier portion. The retina portion processes the photonic image to obtain an image data array of pixels and further process the array of pixels through a window difference network to obtain gradients of the image data. The neural directors of the nonlinear processing portion receive the gradients and generate respective feature vectors, which may have a greater dimensionality than the gradient information, to aid in discrimination between similar patterns in the image data.

Abstract: A neural director is provided which is a neural network constructed with weights that are determined a priori. The neural director receives an input vector X comprising “I” input components “Xi” and generates in response an output vector Y comprising “J” output components. The neural director has an input processing node layer which receives the input vector X and an output processing node layer which generates the output vector Y. The connections between the input and output processing node layers are a unique weighting set w(i,j) that contains an internal representation of a uniform spatial distribution of “J” unit vectors throughout a unit sphere of “I” dimensions. Thus the cosine value between any two adjacent unit vectors is a constant everywhere in the unit sphere.

Abstract: A neural sensor is provided which receives raw input data defining a pattern, such as image or sound data, and generates a classification identifier for the pattern. The neural sensor has a pattern array former which organizes the raw input data into the proper array format. A first order processing section receives the pattern array and generates a first order feature vector illustrative of first order features of the input data. A second order processing section also receives the pattern array and generates at least one second order feature vector illustrative of gradients in the input data. A vector fusion section receives the feature vectors from the first and second order processing sections and generates a single fused feature vector which is provided to a pattern classifier network. The pattern classifier network, in turn, generates a pattern classification for the input data.

Abstract: A dynamic memory processor for time variant pattern recognition and an input data dimensionality reduction is provided having a multi-layer harmonic neural network and a classifier network. The multi-layer harmonic neural network receives a fused feature vector of the pattern to be recognized from a neural sensor and generates output vectors which aid in discrimination between similar patterns. The fused feature vector and each output vector are separately provided to corresponding positional king of the mountain (PKOM) circuits within the classifier network. Each PKOM circuit generates a positional output vector with only one element having a value corresponding to one, the element corresponding to the element of its input vector having the highest contribution. The positional output vectors are mapped into a multidimensional memory space and read by a recognition vector array which generates a plurality of recognition vectors.

Abstract: A color sensor for generating color information defining colors of an image includes an input section, a color processing section, a color comparison section, a color boundary processing section and a memory processing section. The input section includes an array of transducer pairs, each pair defining one of a plurality of pixels. Each transducer pair generates two peak outputs, one for the selected color of each transducer of the pair. A plurality of pixel processors in the color processing section each receives the outputs from one of the transducer pairs. The color processing section generates a color feature vector representative of the brightness of the light incident on the pixels and a color value corresponding to the ratio of outputs from the transducers comprising the transducer pair associated with the pixels. The color boundary processing section generates a plurality of color boundary feature vectors, each representing the difference between the color value for a pixel and its neighboring pixels.

Abstract: A spatial image processor neural network for processing image data to discriminate between first and second spatial configurations of component objects includes a photo transducer input array for converting an input image to pixel data and sending the data to a localized gain network (LGN) module, a parallel memory processor and neuron array for receiving the pixel data and processing the pixel data into component recognition vectors and chaotic oscillators for receiving the recognition vectors and sending feedback data to the LGN module as attention activations. The network further includes a temporal spatial retina for receiving both the pixel data and temporal feedback activations and generating temporal spatial vectors, which are processed by a temporal parallel processor into temporal component recognition vectors. A spatial recognition vector array receives the temporal component recognition vectors and forms an object representation of the first configuration of component objects.

Abstract: A differential hydrophone assembly includes a ceramic hydrophone for disption in a water environment and comprising first and second hemispherical piezoelectric sections, a polarized ceramic member disposed in each section, a positive electrode and a negative electrode disposed in each section, an output conductor extending from each of the electrodes, such that a first pair of the conductors extend from a first of the ceramic members, and a second pair of the conductors extend from a second of the ceramic members, the first pair of conductors being discrete from the second pair of conductors. The assembly further includes a differential summing amplifier for disposition in a second environment removed from the water environment and electrically connected to the first and second pairs of conductors.

Abstract: The present invention relates to an ultra-broadband hydrophone which has major electronic components and one mechanical component. The first electronic component comprises a flux-gate magnetometer for sensing changes in an ambient magnetic field and for creating an electrical signal representative of the ambient magnetic field. The second electronic component comprises an electrical circuit for demodulating the electrical signal and a low pass filter for smoothing the demodulated signal. The mechanical component is a magnetic diaphragm for receiving acoustic pressure waves and for causing changes in the ambient magnetic field in response to the received acoustic pressure waves. In a preferred embodiment, the hydrophone of the present invention also includes a sub-Hertz differential magnetic field winding for generating a differential magnetic field which nulls the ambient magnetic field about the flux-gate magnetometer.

Abstract: A method for testing a neural network, and apparatus for carrying out the method. A first embodiment of the method includes the steps of (a) providing a neural network having a set of connection values, (b) stimulating at least one input of the neural network with an input vector to obtain output signals at an output of the neural network, (c) obtaining a plurality of samples of the output signals, wherein at least one of the plurality of samples is delayed in time from another one of the samples, and wherein at least one of the plurality of samples may represent a difference between two samples, (d) generating an image from the plurality of samples, and (e) comparing the image to a reference image to determine an operational characteristic of the neural network.

Abstract: A nonlinear oscillator (10) includes a neural network (12) having at least one output (12a) for outputting a one dimensional vector. The neural network includes a plurality of layers, including an input layer, an output layer, and at least one hidden layer. Each of the layers includes at least one processing element (PE) that is interconnected to processing elements of adjacent layers. The input layer has an input coupled to the at least one output and includes an analog delay line (14) having a plurality of taps each of which outputs a time-delayed sample of the one dimensional output vector. Each of the taps is connected to each one of the processing elements of the at least one hidden layer for providing a time-delayed sample of the one dimensional output vector thereto.

Abstract: A color sensor for generating color information defining colors of an image includes an input section, a color processing section, a color comparison section, a color boundary processing section and a memory processing section. The input section includes an array of transducer pairs, each pair defining one of a plurality of pixels. Each transducer pair generates two peak outputs, one for the selected color of each transducer of the pair. A plurality of pixel processors in the color processing section each receives the outputs from one of the transducer pairs. The color processing section generates a color feature vector representative of the brightness of the light incident on the pixels and a color value corresponding to the ratio of outputs from the transducers comprising the transducer pair associated with the pixels. The color boundary processing section generates a plurality of color boundary feature vectors, each representing the difference between the color value for a pixel and its neighboring pixels.