Granted Patent B2

US 12,440,147 · App. 17/856,528 · Granted Oct 14, 2025

Inventors: Nilanjan Sarkar (Brentwood, TN); Zhaobo Zheng (Nashville, TN); Amy S. Weitlauf (Nashville, TN); John Staubitz (Nashville, TN); Zachary E. Warren (Nashville, TN); James P. Dieffenderfer (Raleigh, NC)

Assignees: Vanderbilt University; North Carolina State University

A61B5/4076A61B5/02055A61B5/0531A61B5/11A61B5/4803A61B5/6804A61B5/681A61B5/7221A61B5/7267A61B5/74G16H20/70G16H40/63A61B5/02438A61B2562/0204A61B2562/0219A61B2562/04

Abstract

Claims (39)

1. A system for predicting problem behavior in individuals with developmental and behavior disabilities, the system comprising:

a wearable upper body motion sensing device including a plurality of inertial measurement units positioned at known positions relative to each other; and

an electronic controller configured to

receive output signals from each of the plurality of inertial measurement units while the wearable upper body motion sensing device is worn by a subject individual,

model an upper body position of the subject individual by calculating one or more angles for body joints based on the output signals from the plurality of inertial measurement units, wherein the one or more angles include at least one selected from the group consisting of a roll angle, a pitch angle, and a yaw angle,

apply a trained machine-learning model by providing, as input to the trained machine-learning model, an input data set that includes at least one selected from the group consisting of signal data from at least one inertial measurement unit, the upper body position information, and a sequence of upper body position information,

wherein the trained machine-learning model is trained to produce as output, in response to receiving the input data set, an indication of whether a precursor to the problem behavior is detected, and

generate a notification indicating that the precursor to the problem behavior is detected in response to the trained machine-learning model producing an output indicating that the precursor to the problem behavior has been detected.

2. The system of claim 1 , wherein the wearable upper body motion sensing device further includes a wearable upper body garment, wherein each of the plurality of inertial measurement units is affixed to the wearable upper body garments at positions such that, when the wearable upper body garment is worn by the subject individual the plurality of inertial measurement units includes a first inertial measurement unit positioned on a right forearm of the subject individual, a second inertial measurement unit positioned on a left forearm of the subject individual, a third inertial measurement unit positioned on a right upper arm of the subject individual, a fourth inertial measurement unit positioned on a left upper arm of the subject individual, and a fifth inertial measurement unit positioned on a back of the subject individual.

3. The system of claim 2 , wherein the wearable upper body garment includes at least one selected from the group consisting of a shirt, a crewneck sweatshirt, and a hooded sweatshirt.

4. The system of claim 3 , wherein the wearable upper body garment includes the hooded sweatshirt and wherein cabling and electronic circuitry for the wearable upper body motion sensing device is positioned within a hood of the hooded sweatshirt.

5. The system of claim 2 , wherein the wearable upper body garment includes a plurality of pockets positioned at different locations on the wearable upper body garment, wherein each pocket of the plurality of pockets is sized to receive a different individual inertial measurement unit of the plurality of inertial measurement units.

6. The system of claim 5 , wherein each pocket of the plurality of pocket is selectively closeable by a hook-and-loop closure material.

7. The system of claim 1 , wherein the system further includes at least one physiological sensor,

wherein the electronic controller is further configured to receive output signals from the at least one physiological sensor, and

wherein the electronic controller is configured to apply the trained machine-learning model by providing, as input to the trained machine-learning model, an input data set that further includes at least one selected from the group consisting of signal data from the at least one physiological sensor and biometric feature metrics extracted from the signal data from the at least one physiological sensor.

8. The system of claim 7 , wherein the electronic controller is further configured to extract at least one biometric feature from the signal data from the at least one physiological sensor, the at least one biometric feature being selected from the group consisting of a bioimpedance, a heart rate, a skin conductance, and a body temperature.

9. The system of claim 7 , further comprising a wrist worn device including the at least one physiological sensor, wherein the wrist worn device has an appearance of a wristwatch.

10. The system of claim 1 , further comprising a microphone, wherein the electronic controller is configured to receive output signals from the microphone and to extract at least one audio feature from the audio data, the at least one audio feature being selected from the group consisting of individual words, individual phrases, a change in verbal pitch of the subject individual, and a change in verbal volume of the subject individual, and wherein the electronic controller is configured to apply the trained machine-learning model by providing, as the input to the trained machine-learning model, an input data set that further includes at least one selected from the group consisting of the audio data and the at least one audio feature extracted from the audio data.

11. The system of claim 1 , wherein the electronic controller is configured to generate the notification indicating that the precursor to the problem behavior is detected by causing a portable device associated with at least one individual to output the notification, wherein the at least one individual is selected from the group consisting of the subject individual, a parent of the subject individual, a caregiver of the subject individual, a teacher of the subject individual, and a therapist of the subject individual.

12. The system of claim 1 , wherein the electronic controller is configured to generate the notification indicating that the precursor to the problem behavior is detected by causing a portable device associated with the subject individual the notification indicating that the precursor has been detected and identifying a recommended mitigation to prevent the problem behavior from occurring.

13. The system of claim 1 , wherein the electronic controller is further configured to

receive a first user input indicative of an occurrence of a false positive and a second user input indicative of an occurrence of a false negative, wherein the first user input indicates that the machine-learning model determined that a precursor had been detected when not precursor or problem behavior was observed by the user, wherein the second user input indicates that the machine learning model did not determine that the precursor had been detected when precursor or problem behavior was observed by the user, and

retrain the machine-learning model by determining a ground truth label based on the first user input or the second user input and mapping the new ground truth label to signal data from plurality of inertial measurement units captured at a same time as the first user input or the second user input is received.

14. A method of training the machine-learning model used in the system of claim 1 , the method comprising:

equipping the subject individual with the wearable upper body motion sensing device;

administering a behavior assessment while recording signal data from the plurality of inertial measurement units, wherein administering the behavioral assessment includes a controlled sequence of actions in which the subject individual is exposed to stimulus that may provoke the problem behavior and the precursor;

receiving, from a user through a computer-based user device, real-time indications of observed occurrences of the problem behavior and observed occurrences of the precursor while the behavior assessment is administered;

generating a set of training data by

defining a plurality of ground truth labels based on the received real-time indications of the observed occurrences of the problem behavior and the observed occurrences of the precursor, and

mapping each real-time indication to recorded signal data from the plurality of inertial measurement units; and

training the machine-learning model to produce an output corresponding to the ground truth label in response to receiving as input at least one selected from the group consisting of the signal data from at least one inertial measurement unit mapped to the ground truth label, upper body position information determined based on the signal data mapped to the ground truth label, and a sequence of upper body position information determined based on the signal data mapped to the ground truth label.

15. The method of claim 14 , wherein administering the behavior assessment includes administering an interview-informed synthesized contingency analysis (IISCA).

16. The method of claim 14 , wherein training the machine-learning model includes training an individualized machine-learning model based only on ground truth labels and recorded signal data for the subject individual.

17. The method of claim 14 , wherein training the machine-learning model includes training a group machine-learning model based on aggregated ground truth labels and recorded signal data for each of a plurality of subject individuals.

18. The system of claim 1 , wherein the electronic controller is further configured to:

compute a position and an orientation of each body joint of the body joints using forward kinematics based on the one or more angles.

19. The system of claim 1 , wherein the electronic controller is further configured to:

synchronize data modalities having different update frequencies by using the output signals from the plurality of inertial measurement units as a base timeline and interpolating data modalities different from the output signals from the plurality of inertial measurement units into intervals of the base timeline to form unified multimodal data entries.

Assignments (2)

Continuity (2)

References Cited (121)

Cited By (1)

US 12,712,972