Abstract:
The neural engine (20) is a hardware implementation of a neural network for use in real-time systems. The neural engine (20) includes a control circuit (26) and one or more multiply/accumulate circuit (28) includes a parallel/serial arrangement of multiple multiplier/accumulators interconnected with weight storage elements to yield multiple neural weightings and sums in a single clock cycle. A neural processing language is used to program the neural engine (20) through a conventional host personal computer (22). The parallel processing permits very high processing speeds to permit real-time pattern classification capability.
Abstract:
First and second passive sensors (14, 16), which may be mounted on different earth orbiting satellites, provide relative azimuth and elevation coordinates to sensed objects (A, B, C) such as hostile missiles. Minimum and maximum possible ranges to the objects (A, B, C) along lines-of-sight (18a, 18b, 18c) from the first sensor (14) are predetermined, and used to calculate "range lines" (24, 26, 28) which are coincident with the lines-of-sight (18a, 18b, 18c) and extend from the respective minimum to maximum ranges respectively. The range lines (24, 26, 28) are transformed into the field of view of the second sensor (16), and matched to the azimuth and elevation coordinates of the respective objects (A, B, C) using a basic feasible solution (greedy) or global optimization algorithm. The approximate points of intersection of lines-of-sight (20a, 20b, 20c) from the second sensor (16) to the objects (A, B, C) and matched range lines (24, 26, 28), are calculated, and transformed into coordinates relative to a reference point (10) such as the center of the earth (12). The calculations for the individual objects (A, B, C) may be performed simultaneously using parallel processors (44a, 44b, 44c) in a single instruction stream - multiple data stream (SIMD) or similar computing arrangement (40).
Abstract:
First and second passive sensors (14, 16), which may be mounted on different earth orbiting satellites, provide relative azimuth and elevation coordinates to sensed objects (A, B, C) such as hostile missiles. Minimum and maximum possible ranges to the objects (A, B, C) along lines-of-sight (18a, 18b, 18c) from the first sensor (14) are predetermined, and used to calculate 'range lines' (24, 26, 28) which are coincident with the lines-of-sight (18a, 18b, 18c) and extend from the respective minimum to maximum ranges respectively. The range lines (24, 26, 28) are transformed into the field of view of the second sensor (16), and matched to the azimuth and elevation coordinates of the respective objects (A, B, C) using a basic feasible solution (greedy) or global optimization algorithm. The approximate points of intersection of lines-of-sight (20a, 20b, 20c) from the second sensor (16) to the objects (A, B, C) and matched range lines (24, 26, 28), are calculated, and transformed into coordinates relative to a reference point (10) such as the center of the earth (12). The calculations for the individual objects (A, B, C) may be performed simultaneously using parallel processors (44a, 44b, 44c) in a single instruction stream - multiple data stream (SIMD) or similar computing arrangement (40).
Abstract:
Des premier et second détecteurs passifs (14, 16), pouvant être montés sur des satellites placés sur différentes orbites terrestres, fournissent des coordonnées relatives d'azimut et d'élévation à des objets détectés (A, B, C) tels que des missiles hostiles. Des distances possibles minimum et maximum aux objets (A, B, C) le long de lignes de visée (18a, 18b, 18c) sont prédéterminées à partir du premier détecteur (14), et sont utilisées pour calculer "des lignes de distance" (24, 26, 28) qui coïncident avec les lignes de visée (18a, 18b, 18c) et s'étendent depuis les distances respectives minimum à maximum, respectivement. Les lignes de distance (24, 26, 28) sont transformées en champ de vision du second détecteur (16) pour correspondre aux coordonnées d'azimut et d'élévation des objets respectifs (A, B, C) en utilisant une solution faisable de base (avide) ou un algorithme d'optimalisation globale. Les points approximatifs d'intersection des lignes de visée (20a, 20b, 20c) allant du second détecteur (16) aux objets (A, B, C) et des lignes de distances appariées (24, 26, 28) sont calculés et transformés en coordonnées par rapport à un point de référence (10) tel que le centre de la terre (12). Les calculs pour les objets individuels (A, B, C) peuvent être effectués simultanément en utilisant des processeurs parallèles (44a, 44b, 44c) dans un train de mono-instruction/multidonnées (SIMD) ou un agencement de calcul similaire (40).