US 12,460,954 B2
Recognition apparatus, recognition method, and non-transitory computer-readable storage medium
Yasunobu Yamauchi, Yokohama Kanagawa (JP)
Assigned to Kabushiki Kaisha Toshiba, Kawasaki (JP)
Filed by KABUSHIKI KAISHA TOSHIBA, Tokyo (JP)
Filed on Aug. 31, 2022, as Appl. No. 17/823,531.
Claims priority of application No. 2022-009531 (JP), filed on Jan. 25, 2022.
Prior Publication US 2023/0236047 A1, Jul. 27, 2023
Int. Cl. G01D 21/02 (2006.01); G06V 10/44 (2022.01); G06V 10/764 (2022.01); G06V 10/80 (2022.01)
CPC G01D 21/02 (2013.01) [G06V 10/443 (2022.01); G06V 10/764 (2022.01); G06V 10/806 (2022.01)] 20 Claims
OG exemplary drawing
 
1. A recognition apparatus comprising:
processing circuitry configured to:
receive sensor data from a sensor;
generate, using a feature quantity extracting neural network, a first feature quantity exhibiting a feature of the sensor data based on the sensor data;
convert, using a feature quantity converting neural network, the first feature quantity into a second feature quantity exhibiting a feature contributing to identification of a class of the sensor data;
generate, using an adjustment neural network, a generic feature quantity having a same number of filters as the second feature quantity based on the first feature quantity;
calculate a feature quantity significance by multiplying the generic feature quantity with the second feature quantity;
generate a significant feature quantity exhibiting a feature that is significant in the identification of the class based on a cross-correlation between the first feature quantity and the second feature quantity, by multiplying the generic feature quantity with the feature quantity significance;
generate an integrated feature quantity considering features of the first feature quantity and the second feature quantity, based on the second feature quantity and the significant feature quantity; and
identify, using a recurrent neural network, the class based on the integrated feature quantity.