Title:
TIME SERIES SENSOR DATA PROCESSING APPARATUS AND TIME SERIES SENSOR DATA PROCESSING METHOD
Document Type and Number:
Japanese Patent JP2022007313
Kind Code:
A
Abstract:
To provide an apparatus which can determine whether a data pattern is normal or abnormal in accordance with its occurrence timing even in the case that the data pattern is a specific data pattern.SOLUTION: A CPU 101 of an MPU 100 includes a frame division unit 31, an FFT processing unit 32, a time information addition unit 33, a neural network inference processing unit 151, and a result determination processing unit 152. The frame division unit 31 divides time series sensor data to generate a plurality of frames so that each of the frames has a prescribed frame size. The time information addition unit 33 adds a value indicative of time information corresponding to each frame of the frames subjected to FFT processing by the FFT processing unit 32, to a frequency spectrum of the frame. The neural network inference processing unit 151 uses a neural network parameter 140a stored in a memory 102 to execute neural network inference processing, and the result determination processing unit 152 executes result determination processing on the basis of a result of the neural network inference processing.SELECTED DRAWING: Figure 2
Inventors:
SHIMOBEPPU MASAYUKI
Application Number:
JP2020110215A
Publication Date:
January 13, 2022
Filing Date:
June 26, 2020
Export Citation:
Assignee:
RENESAS ELECTRONICS CORP
International Classes:
G06N3/08; G01H17/00; G06N20/00
Attorney, Agent or Firm:
Tsutsui International Patent Office
Previous Patent: Double glazing
Next Patent: Vibration actuators, lens barrels and electronic devices
Next Patent: Vibration actuators, lens barrels and electronic devices