4 results about "Information fusion" patented technology

The invention discloses an identity recognition method based on multi-modal information. The identity recognition method comprises the following steps: 1, making a multi-modal video data set with a label; 2, respectively constructing and training a face detection model and a head detection model; 3, constructing and training a feature extraction model of the face, the head and the sound; 4, performing feature extraction on the face, head and sound information through the trained feature extraction models; 5, constructing and training a classification model to respectively classify the three extracted features; 6, performing result prediction by using the three features through the classification model; 7, performing information fusion on the classification result according to the formulated multi-modal information fusion strategy; 8, arranging the fused result and then outputting an identity recognition result. Based on the identity recognition network model based on multi-modal information provided by the invention, the identity recognition network model has a wide application prospect in the fields of human-computer interaction, information security, security monitoring and the like.

The invention discloses a multi-modal target detection method and system suitable for modal missing. The method comprises the following steps: S1, training a neural network unit by using multi-modal data in a data set; inputting all modal data of the data set into the trained neural network unit for detection, and storing a detection result; s2, extracting modal data of other dimensions; generating a pseudo-modal feature tensor by using the generative network unit, splicing, and inputting into the attention network unit, the information fusion unit and the discrimination network unit until the training of the generative network unit is completed; and S3, collecting modal data in real time through data collection equipment, and generating the type and identifier of the target by using the trained neural network unit. According to the method, on the premise that noise introduction and feature information loss are avoided, missing modal data are virtually generated, the calculation amount and complexity of the model are reduced to a great extent, and the representation consistency of generated modals can be improved.

The invention relates to the technical field of robots, and in particular relates to a movable teaching platform. The movable teaching platform comprises a platform body, a plurality of sensors which are integrated on the platform body and are used for collecting road condition information and motion information of the platform body, a processor which is integrated on the platform body, is connected with the plurality of sensors, and is used for carrying out information fusion on the road condition information and the motion information to obtain an information fusion result, a controller which is integrated on the platform body, connected with the processor and used for controlling a braking module according to the information fusion result, and the braking module which is integrated on the platform body, connected with the controller and used for controlling the motion state of the platform body under the control of the controller. According to the embodiment of the invention, the problems that the movable range is narrow, the integration level is low, and the teaching platform is poor in use universality due to the fact that the teaching platform can only be used by some people in the prior art can be solved.