| CPC G06V 30/293 (2022.01) [G06F 18/217 (2023.01); G06N 3/044 (2023.01); G06N 3/084 (2013.01); G06V 10/764 (2022.01); G06V 30/1916 (2022.01); G10L 15/063 (2013.01); G06V 30/287 (2022.01)] | 17 Claims |

|
1. A feature extraction system based on neural network optimization by gradient filtering, comprising:
an information acquisition device, configured to acquire and input continuous input information to a feature extraction device, wherein the continuous input information comprises one or more selected from the group consisting of picture information, voice information and text information;
the feature extraction device, configured to construct feature extraction networks for different input information, iteratively update gradient filter parameters of the feature extraction networks separately in combination with corresponding training task queues, obtain and store optimized feature extraction networks for different input information, and call a corresponding optimized feature extraction network to perform continuous feature extraction according to a class of the continuous input information acquired by the information acquisition device to obtain a continuous feature corresponding to the continuous input information;
an online optimization device, configured to implement online continuous updating of the feature extraction networks during the continuous feature extraction of the continuous input information; and
a feature output device, configured to output the continuous feature corresponding to the continuous input information.
|