US 12,340,303 B2
Computer-implemented method and device for generating frequency component vector of time-series data
Ryunosuke Daido, Hamamatsu (JP); and Kanru Hua, Chiyoda-ku (JP)
Assigned to YAMAHA CORPORATION, Hamamatsu (JP)
Filed by YAMAHA CORPORATION, Hamamatsu (JP)
Filed on Feb. 9, 2021, as Appl. No. 17/171,453.
Application 17/171,453 is a continuation of application No. PCT/JP2019/031380, filed on Aug. 8, 2019.
Claims priority of application No. 2018-151748 (JP), filed on Aug. 10, 2018; and application No. 2018-169304 (JP), filed on Sep. 11, 2018.
Prior Publication US 2021/0166128 A1, Jun. 3, 2021
Int. Cl. G06N 3/08 (2023.01); G06N 3/045 (2023.01)
CPC G06N 3/08 (2013.01) [G06N 3/045 (2023.01)] 12 Claims
OG exemplary drawing
 
1. A computer-implemented method for generating a frequency component vector of time series data in each unit step of a plurality of unit steps, by executing a first process and a second process in each unit step, wherein:
the first process comprises the steps of:
receiving first data; and
processing the first data using a first neural network to generate intermediate data as an alternative representation of the first data; and
the second process comprises the steps of:
receiving the generated intermediate data; and
generating, as the frequency component vector and using a second neural network including a plurality of distinct neural networks, a plurality of component values corresponding to a plurality of frequency bands based on the generated intermediate data such that:
a first component value corresponding to a first frequency band is generated using a first distinct neural network among the plurality of distinct neural networks of the second neural network based on the generated intermediate data;
a second component value corresponding to a second frequency band different from the first frequency band is generated using a second distinct neural network among the plurality of distinct neural networks of the second neural network based on the generated intermediate data and the generated first component value corresponding to the first frequency band generated using the first distinct neural network among the plurality of distinct neural networks of the second neural network; and
a third component value corresponding to a third frequency band different from the second frequency band is generated using a third distinct neural network among the plurality of distinct neural networks of the second neural network based on the generated intermediate data, the generated first component value corresponding to the first frequency band generated using the first distinct neural network among the plurality of distinct neural networks of the second neural network, and the generated second component value corresponding to the second frequency band generated using the second distinct neural network among the plurality of distinct neural networks of the second neural network.