Browse Prior Art Database

A Method for Detecting a Stressful Event Based on Electrocardiogram (ECG) and Respiratory Inductance Plethysmography (RIP) Disclosure Number: IPCOM000248741D
Publication Date: 2017-Jan-04
Document File: 7 page(s) / 362K

Publishing Venue

The Prior Art Database


Chronic and repeated exposure to mental stress has become one of the major factors causing negative consequences to people’s health and has been shown to be linked with a higher occurrence rate of psychological disorder such as depression and physical illness such as cardiovascular diseases. Therefore it is of critical importance to design methods and create tools to help people with their stress management as part of the preventive measures to reduce the likelihood of them slipping into those serious medical conditions. As our initial step toward this overall goal, we study the possible association between the certain physiological parameters which are either provided by or can be derived from data collected by wearable sensors and mental stress that a subject is exposed to in a lab controlled environment. Electrocardiogram (ECG) and respiratory inductance plethysmography (RIP) sensors were applied to subjects as they went through a series activities including stress tests and guided meditation. Instead of using these raw physiological signals in studying the effect exerted by mental stress, most of previous works attempt to take advantage of certain derived but physiological meaningful features from these raw ECG and RIP signals to characterize individuals' stress response. Most segmented all the derived time-series features into small non-overlapping windows and calculate the average for each window before training support vector machine and decision tree models for mental stress detection based on data collected from all the subjects. Therefore, models obtained largely ignore the huge between-subject variability in subjects’ physiological response to the stressor. None of these works attempted to build a truly individualized model that only uses data from an individual. Furthermore, none of these works address the issue of detecting the onset of mental stress until it lasts for a specified period of time which is more directly related to stress management. Our study is the first to propose a framework that builds a truly individualized model for detecting the onset of mental stress.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 24% of the total text.


A Method for Detecting a Stressful Event Based on Electrocardiogram (ECG) and Respiratory Inductance Plethysmography (RIP)


- Chronic or repeated exposures to psychological stress is linked with a higher risk of physical illness such as diabetes as well as affective disorders such as depression [1-4]. - ECG and RIP sensor signals may help discover unique patterns demonstrated in individual’s physiological response to stressors, detect or predict stress episodes and provide insights on appropriate time for stress-reducing interventions [5]. - Currently there still lacks effective models to capture unique stress patterns for an individual through continuous monitoring of physiological parameters. State-of-the-art

- Current methods for detecting mental stress use raw signals and certain derived features (such as time and frequency domain parameters) by segmenting them into non-overlapping small windows. Most models were built using Support Vector Machine (SVM), decision tree, etc., on data collected from all the subjects to detect if a stressor is present in a small time window. Attempts have also been made to associate these features with subjects’ perceived level of stress on a group level [6] versus individual level. - However, data-driven approach has not been applied to extract features from those parameters. Measures have not been taken to deal with the potential noise in those features or control for the specific context the subjects are in. Nor has any effort been made toward discovering the unique patterns demonstrated in each of individual’s physiological response to the stressor. Methods used to model the difference in stressor vs. non-stressor responses cannot be directly applied to a few examples of stress response collected from each of those individuals. What is new in this invention

- A method that can detect the time point up to which an individual has been exposed to a particular stressor for a specified period of time by learning robust to noise local representations of each subject’s physiological responses and building models on those representations. - The method can take into account the huge variability between different individuals by building one model per subject and discover the unique pattern demonstrated in each of individual’s physiological response to


stressor. A method and system that detects a stressful event that has been going on for a specified period of time for a subject by discovering the unique pattern exhibited in the subject’s physiological response to the stressor.

The core of the invention is the derivation of the following analytics methods: - A method that takes as input physiological signals collected from multiple sources and a few examples of the target stressful event going on for a period of time; - A method that uses a data-driven approach to extract features from the time-varying physiological parameters. - A method that models the continuous local changes of signals and learns...