US 11,720,814 B2
Method and system for classifying time-series data
Yilin Shen, Sunnyvale, CA (US); Yue Deng, Sunnyvale, CA (US); and Hongxia Jin, Mountain View, CA (US)
Assigned to SAMSUNG ELECTRONICS CO., LTD., Suwon-si (KR)
Filed by Samsung Electronics Co., Ltd., Suwon-si (KR)
Filed on Dec. 27, 2018, as Appl. No. 16/234,433.
Claims priority of provisional application 62/612,203, filed on Dec. 29, 2017.
Prior Publication US 2019/0205786 A1, Jul. 4, 2019
Int. Cl. G06N 20/00 (2019.01); G06F 18/24 (2023.01); G06V 10/764 (2022.01); G06N 3/049 (2023.01)
CPC G06N 20/00 (2019.01) [G06F 18/24 (2023.01); G06V 10/764 (2022.01); G06N 3/049 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A recognition method, comprising:
receiving, by an electronic device, an input comprising data over a first window size;
training, by the electronic device, a classification model using training pruned feature vectors from pruned feature vectors of training data;
performing testing processing, by the electronic device, of the trained classification model using testing pruned feature vectors from pruned feature vectors of the data to obtain classification model accuracy, wherein the classification model classifies the input based on comparison of warping distance of the input with a pruning threshold; and
performing testing processing, by the electronic device, of a baseline classification model using the testing pruned feature vectors to obtain baseline classification model accuracy, wherein the pruning threshold is based on the classification model accuracy and the baseline classification model accuracy.