US 12,230,366 B2
System and method for fleet driver biometric tracking
Austin Green, Fort Worth, TX (US); and Steven Kastelic, Broken Arrow, OK (US)
Assigned to BlyncSync Technologies, LLC, Broken Arrow, OK (US)
Filed by BlyncSync Technologies, LLC, Broken Arrow, OK (US)
Filed on Oct. 4, 2022, as Appl. No. 17/959,355.
Application 17/959,355 is a continuation of application No. 15/615,005, filed on Jun. 6, 2017, granted, now 11,462,301, issued on Oct. 4, 2022.
Claims priority of provisional application 62/346,796, filed on Jun. 7, 2016.
Prior Publication US 2023/0142766 A1, May 11, 2023
This patent is subject to a terminal disclaimer.
Int. Cl. G16B 40/20 (2019.01); B60W 40/09 (2012.01); B60W 50/14 (2020.01); G06F 16/435 (2019.01); G06N 3/02 (2006.01); G16B 40/00 (2019.01); G16H 40/67 (2018.01); G16H 50/20 (2018.01)
CPC G16B 40/20 (2019.02) [B60W 40/09 (2013.01); B60W 50/14 (2013.01); G06F 16/435 (2019.01); G06N 3/02 (2013.01); G16B 40/00 (2019.02); G16H 40/67 (2018.01); G16H 50/20 (2018.01)] 16 Claims
OG exemplary drawing
 
1. A method for biometric tracking, comprising:
receiving, using one or more wearable devices, a plurality of biometric data including blink rate data, head acceleration data, and sleep data, wherein the sleep data includes REM sleep data;
creating and updating over time an electronic log, wherein the electronic log includes a plurality of states of a user each mapped to one or more time periods and each mapped to one or more of the plurality of biometric data;
inputting, into a predictive engine, biometric data selected from the plurality of biometric data, wherein the predictive engine is a machine learning model;
training the predictive engine using the biometric data;
retraining the predictive engine to provide for an adjusted a predicted behavior pattern of the predictive engine based on the sleep data received over time, including adjusting an under-prediction or an over-prediction by gradually increasing or decreasing a predicted number of awake hours at periodic intervals, including:
adjusting the under-prediction by the predictive engine by gradually increasing the predicted number of awake hours at the periodic intervals, wherein the under-prediction indicates that the predicted number of awake hours for the user is predicted based on tracked REM sleep hours for the user and is predicted less than a number of actual awake hours of the user;
adjusting the over-prediction by the predictive engine by gradually decreasing the predicted number of awake hours at the periodic intervals, wherein the over-prediction indicates that the predicted number of awake hours for the user is predicted based on the tracked REM sleep hours for the user and is predicted greater than the number of actual awake hours of the user; and
modifying weights of the predictive engine at the periodic intervals according to the adjusted predicted behavior pattern via backpropagation of errors through the predictive engine; and
outputting the adjusted predicted behavior pattern using the predictive engine and the modified weights, including:
providing biometric inputs to the predictive engine;
manipulating the inputs by the modified weights; and
applying an activation function to the manipulated inputs.