System and method for assessing physiological state
Inventors
BARNETT, Jennifer Helen • CORMACK, Francesca Kathleen • TAPTIKLIS, Nicholas Theodore • SU, Merina Tong
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-12347563-B2
Publication Date
2025-07-01
Expiration Date
Abstract
A system for assessing the physiological state of a subject, comprising: a task delivery module configured to communicate to a subject at least two sets of information, each set of information relating to a cognitive task requiring a spoken response from the subject; a response detection module configured to record the respective spoken responses from the subject as an audio signal, the response detection module comprising a microphone; an analysis module configured to analyze the audio signals corresponding to the respective spoken responses recorded by the response detection module to determine from the respective spoken responses one or more characteristics indicative of the physiological state of the subject, compare said characteristics from the respective spoken responses, and determine the physiological state of the subject based on said comparison.
Core Innovation
The invention provides a mobile computer device configured to predict a physiological state of a subject based on spoken responses collected during a sequence of cognitive tasks. The physiological state includes a level of pain, a level of alertness, fatigue or sedation, or a level of stress or anxiety experienced by the subject. The device includes a user interface device and a microphone to communicate instructions and receive audio signals comprising spoken responses to the tasks.
The device stores a plurality of cognitive tasks requiring a spoken response, where each task is associated with a cognitive load and with sets of instructions for performing the cognitive tasks. The one or more processors generate a sequence of successive cognitive tasks having systematically varying cognitive load, then control the user interface device to communicate instructions for each cognitive task and control the microphone to receive spoken responses corresponding to different cognitive loads. The processor also receives task performance signals related to each different cognitive load.
For each successive cognitive task, the processors analyze audio signals and task performance signals to determine speech characteristics of the spoken responses by detecting portions of the spoken responses that correspond to different cognitive loads. The processors calculate values associated with different speech characteristics comprising at least two of pitch, intensity, formant frequencies, glottal flow, speech duration, speech rate, and voice quality, extract acoustic features, and calculate delta values for the acoustic features from higher cognitive load tasks to lower cognitive load tasks. By comparing values to delta values, the processors create speech characteristics related to the current physiological state while cancelling speech characteristics related to task conditions but not the speech characteristics related to the current physiological state.
Using AI/ML trained on training data relating human generated audio signals to a physiological state, the processors perform a prediction of the current physiological state based at least in part on the speech characteristics related to the current physiological state within portions of the audio signals. The prediction is output as a determination of the physiological state level of pain, alertness, fatigue or sedation, or stress or anxiety, and is communicated to at least one of the subject via the user interface device or a clinical team, medical health record, or pharmacist.
Claims Coverage
The document contains one independent claim. Its core inventive structure comprises administering a systematically varying sequence of cognitive tasks with spoken responses, extracting and delta-canceling speech characteristics tied to physiological state, and using AI/ML trained on human audio-to-physiology data to predict and communicate a physiological-state level.
Physiological-state prediction from spoken cognitive tasks on a mobile device
A mobile computer device configured to predict a physiological state of a subject, including a level of pain, a level of alertness, fatigue or sedation, or a level of stress or anxiety experienced by the subject, using cognitive tasks requiring spoken responses, microphone-received audio signals, and AI/ML trained on human generated audio signals to a physiological state.
Systematically varying cognitive-load task sequence with spoken responses and task performance signals
The one or more processors generate a sequence comprising respective successive cognitive tasks selected to have systematically varying different associated cognitive load, control the user interface device to communicate instructions for performing the cognitive tasks in the sequence, control the microphone to receive audio signals comprising spoken responses related to different cognitive loads, store the audio signals, and receive task performance signals related to each different cognitive load.
Speech-characteristic extraction per cognitive load with detected portions, acoustic features, and delta values
For each successive cognitive task, the processors analyze audio signals and task performance signals to determine speech characteristics by detecting portions of the spoken responses that correspond to different cognitive loads, calculate values associated with speech characteristics comprising at least two of pitch, intensity, formant frequencies, glottal flow, speech duration, speech rate, and voice quality, extract acoustic features corresponding to the different speech characteristics, and calculate delta values for each acoustic feature from higher cognitive load tasks to lower cognitive load tasks.
Delta-based cancellation of task-condition speech characteristics to isolate current physiological state
The processors compare the respective values to the delta values to determine speech characteristics related to a current physiological state of the subject, thereby cancelling out speech characteristics related to task conditions but not the speech characteristics related to the current physiological state, to create a comparison.
AI/ML prediction and communication of physiological-state determination
The processors perform a prediction of the current physiological state of the subject based at least in part on the speech characteristics related to the current physiological state within portions of the audio signals using AI/ML based on training data relating human generated audio signals to a physiological state, and output and communicate a determination of the physiological state level to at least one of the subject via the user interface device or a clinical team, medical health record, or pharmacist.
The independent claim covers a mobile-device framework that administers a systematically varying cognitive-load sequence, detects and analyzes spoken-response portions to compute speech characteristics, computes delta values across cognitive loads to cancel task-condition effects, and applies AI/ML trained on human audio-to-physiology data to predict and communicate a physiological-state level.
Stated Advantages
Cancels speech characteristics related to task conditions but not the speech characteristics related to the current physiological state.
Communicates the physiological-state determination to at least one of the subject, a clinical team, a medical health record, or a pharmacist.
Documented Applications
Remote monitoring.
Repeat prescription decision support.
Post-operative discharge readiness.
Intervention effectiveness monitoring.
Safety control for high-risk jobs/activities.
Consumer self-monitoring.
Clinical trials.
Interested in licensing this patent?