Signal processing unit for classifying objects on the basis of signals from sensors

Claim 2

A signal processing unit for classifying objects comprising: a plurality of sensors responding to different physical quantities detected at an object to be classified and providing sensor information, pairs of first and second neural networks, each of said pairs being associated with a respective one of said sensors and arranged to receive said information from said one of said sensors, said first neural network of each pair being adapted and trained to process image information of said object to be classified, said second neural network of each pair being adapted and trained to process spectral information of said object to be classified, said first neural networks providing detection, identification and movement information derived from said image information and specific for the respective one of said sensors, said second neural networks providing detection, identification and movement information derived from said spectral information and specific for the respective ones of said sensors, said information, combined, representing a feature vector, a third neural network to which said feature vector is applied, said third neural network being adapted and trained to determine the associations of said feature vector of said detection, identification and movement information, to provide association information, and identifying and classifying means for identifying and classifying said object, said identifying and classifying means being arranged to receive said association information and to provide final identification information.

Claim 3

A signal processing unit as claimed in claim 1, wherein said identifying and classifying means is arranged to receive also said feature vector.

Claim 4

A signal processing unit as claimed in claim 1, wherein said identifying and classifying means comprise a fourth neural network, to which said association information is applied, said fourth neural network being trained to identify and classify said object to provide identification and classification information.

Claim 5

A signal processing unit as claimed in claim 3, wherein said fourth neural network is arranged to receive also said feature vector.

Claim 6

A signal processing unit as claimed in claim 3, and further comprising an expert system means, said expert system means having stored therein knowledge about data and facts relating to potential objects, said expert system being arranged to receive said association information from said third neural network and said identification and classification information from said fourth neural network to provide therefrom and from said knowledge said final identification information.

Claim 7

A signal processing unit as claimed in claim 1, wherein said sensors comprise a radar sensor, a laser sensor and an infrared sensor.

Claim 8

A signal processing unit as claimed in claim 1, wherein said sensors are adapted to provide picture information of said object.

Claim 9

A signal processing unit as claimed in claim 7, wherein a first neural network associated with one of said sensors processes image information to provide detection and identification information specific for said one sensor.

Claim 10

A signal processing unit as claimed in claim 1, wherein said neural networks are multi-layer feedforward networks.

Claim 11

A signal processing unit for classifying objects comprising: a plurality of sensors responding to different physical quantities detected at an object to be classified and providing sensor information, pairs of first and second multi-layer feedforward neural networks, each of said pairs being associated with a respective one of said sensors and arranged to receive said information from said one of said sensors, said first neural network of each pair being adapted and trained to process image information of said object to be classified, said second neural network of each pair being adapted and trained to process spectral information of said object to be classified for the detection and identification of potential targets, said first neural networks providing detection, identification and movement information derived from said image information and specific for the respective ones of said sensors, said second neural networks providing detection, identification and movement information derived from said spectral information and specific for the respective ones of said sensors, said information, combined, representing a feature vector, a third neural network to which said feature vector is applied, said third neural network being adapted and trained to determine the associations of said feature vector of said detection, identification and movement information, to provide association information, a fourth neural network, to which said association information and said feature vector are applied, said fourth neural network being trained to identify and classify said object to provide identification and classification information, and an expert system means, said expert system means having stored therein knowledge about data and facts relating to potential objects, said expert system means being arranged to receive said association information from said third neural network and said identification and classification information from said fourth neural network to provide therefrom and from said knowledge final identification information.

Claim 12

A signal processing unit for classifying objects comprising: a plurality of sensors responding to different physical quantities detected at an object to be classified and providing sensor information, pairs of first and second multi-layer feedforward neural networks, each of said pairs being associated with a respective one of said sensors and arranged to receive said information from said one of said sensors, said first neural network of each pair being adapted and trained to process image information of said object to be classified, said second neural network of each pair being adapted and trained to process spectral information of said object to be classified for the detection and identification of potential targets, said first neural networks providing detection, identification and movement information derived from said image information and specific for the respective ones of said sensors, said second neural networks providing detection, identification and movement information derived from said spectral information and specific for the respective ones of said sensors, said information, combined, representing a feature vector, a third neural network to which said feature vector is applied, said third neural network being adapted and trained to determine the associations of said feature vector of said detection, identification and movement information, to provide association information, a fourth neural network, to which said association information and said feature vector are applied, said fourth neural network being trained to identify and classify said object to provide identification and classification information, and an expert system means, said expert system means having stored therein knowledge about data and facts relating to potential objects, said expert system means being arranged to receive said association information from said third neural network and said identification and classification information from said fourth neural network to provide therefrom and from said knowledge final identification information.