Predicting wellness of a user with monitoring from portable monitoring devices
Inventors
Rezai, Ali • Finomore, Victor • D'HAESE, Pierre • MARSH, Clay
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Assignees
MemberWest Virginia UniversityWest Virginia UniversityWest Virginia University is a public R1 research institution offering diverse undergraduate and graduate programs across science, engineering, business, creative arts, and media. The university emphasizes experiential learning, research, and innovation, with notable strengths in academic program development, research excellence, community engagement, and a commitment to affordability, career readiness, and student success. WVU supports a vibrant campus environment, industry partnerships, and impactful scholarship, preparing students for careers through hands-on education, research, and applied learning.
West Virginia University is a public R1 research institution offering diverse undergraduate and graduate programs across science, engineering, business, creative arts, and media. The university emphasizes experiential learning, research, and innovation, with notable strengths in academic program development, research excellence, community engagement, and a commitment to affordability, career readiness, and student success. WVU supports a vibrant campus environment, industry partnerships, and impactful scholarship, preparing students for careers through hands-on education, research, and applied learning.
Publication Number
US-12544012-B2
Publication Date
2026-02-10
Expiration Date
Abstract
Systems and methods are provided for monitoring a wellness of a user. A wellness-relevant parameter representing the user is monitored at a portable device over a defined period to produce a time series for the wellness-relevant parameter. A first set and a second set of either cognitive assessment data or psychosocial assessment data are obtained for the user at respective first and second times in the defined period. A value is assigned to the user via a predictive model according to the time series for the wellness-relevant parameter, the first set of either cognitive assessment data or psychosocial assessment data, and the second set of either cognitive assessment data or psychosocial assessment data.
Core Innovation
A method for monitoring a wellness of a user assigns a value representing a wellness of the user based on physiological time-series data and cognitive and psychosocial assessment data. The physiological input is obtained by monitoring one of a heart rate or a heart-rate variability of the user at a heart rate sensor over a defined period to produce a time series.
The assigning uses a recurrent neural network that receives the time series representation derived by a wavelet decomposition and the cognitive and psychosocial assessment data sets. A wavelet decomposition is performed on the time series to provide a two-dimension array of wavelet coefficients across first and second variables, and a center of mass of the two-dimensional array is generated as a first representative value for the first variable and a second representative value for the second variable.
The value is assigned according to at least the first and second representative values. The method measures an outcome associated with the user and compares the measured outcome to the value assigned to the user by the recurrent neural network, and a parameter associated with the recurrent neural network is changed according to the comparison by generating a reward for a reinforcement learning process based on a similarity of the measured outcome to the value assigned.
Claims Coverage
Two independent claims are present: clm-00001 (method) and clm-00008 (system). Both include a recurrent neural network assigning a wellness-related value or index using wavelet-decomposed features derived from a two-dimensional array of wavelet coefficients, together with cognitive and psychosocial assessment data, and both use reinforcement learning to refine a decision threshold or network parameters based on measured outcomes.
Recurrent neural network wellness assignment with wavelet center-of-mass features
Assigning a value representing wellness to the user via a recurrent neural network according to the time series for one of heart rate or heart-rate variability, a first set and a second set of one of cognitive assessment data and psychosocial assessment data, wherein the assigning comprises performing a wavelet decomposition to provide a two-dimension array of wavelet coefficients and generating a center of mass of the two-dimensional array as first and second representative values.
Reinforcement learning parameter or threshold refinement from measured outcome similarity
Measuring an outcome associated with the user, comparing the measured outcome to the value assigned to the user via the recurrent neural network, and changing a parameter associated with the recurrent neural network by generating a reward for a reinforcement learning process based on a similarity of the measured outcome to the value assigned.
Wavelet feature extraction with representative values for recurrent neural network
A feature extractor that performs a wavelet decomposition on the time series to provide a two-dimensional array of wavelet coefficients across first and second variables and generates a center of mass of the two-dimensional array as a first representative value for the first variable and a second representative value for the second variable.
Reinforcement learning model refining decision threshold for categorical wellness parameter
A reinforcement learning model that continuously refines a decision threshold applied to the index representing the user to assign a categorical parameter representing a wellness of the user.
The independent claims center on wavelet decomposition of heart rate or heart-rate variability time series into a two-dimensional array of wavelet coefficients, generation of a center-of-mass representative feature pair, and recurrent neural network assignment of a wellness-related value or index using those features together with first and second cognitive and psychosocial assessment data, followed by reinforcement learning to refine network behavior or a decision threshold based on measured outcomes and similarity.
Stated Advantages
Not explicitly described in patent.
Documented Applications
Not explicitly described in patent.
Interested in licensing this patent?