Title:
METHOD FOR IDENTIFYING INDIVIDUAL BY MACHINE LEARNING OF FEATURE AT FLICK INPUT TIME
Document Type and Number:
Japanese Patent JP2018005868
Kind Code:
A
Abstract:
PROBLEM TO BE SOLVED: To identify an individual based on features at character input time in flick input on a smartphone.SOLUTION: At flick input time, the following elements are acquired: movement distances from a sensor mounted on a smartphone to X and Y axis directions; movement distances to X and Y axis directions in a unit time; a time when a user touched a touch screen; acceleration and angular speed in each of X, Y, Z axes in a moment when a user touched a screen; and acceleration and angular speed in each of X, Y, Z axes in a moment when a user's finger separated from the screen. Then, these elements are subjected to machine learning. Plural subjects who are comparison objects are prepared, and each subject performs input by almost 50 times in all characters which can be input. All input results are classified for every character, and all input results are learned by a self-organization map to calculate an area of each subject. Thereafter, a certain subject inputs a certain character, and the input character is compared with a learned self-organization map corresponding to the input character, therefore it is determined which subject performed input.
Inventors:
UDA TAKAYA
NOHARA TAKUMI
NOHARA TAKUMI
Application Number:
JP2016141855A
Publication Date:
January 11, 2018
Filing Date:
June 30, 2016
Export Citation:
Assignee:
UDA TAKAYA
International Classes:
G06F3/0488; G06F3/023; H03M11/04