MONITORING WELLNESS USING A WIRELESS HANDHELD DEVICE

1. A method for detecting a subject's stress level associated with an activity, comprising:

connecting the subject to a sensor that senses a value of a biometric;

during the activity,

(a) repeatedly sensing the value of the biometric over each of a plurality of time windows; and

(b) computing, for each time window, a deviation in the sensed values of the biometric; and

detecting the stress level based on the computed deviations.

2. A method as in claim 1, wherein the value of the biometric is a skin temperature measurement.

3. A method as in claim 1, wherein the deviation for each window is computed as the difference between the maximum deviation in that time window of the sensed value of the biometric from a reference value of the biometric and the minimum deviation in that time window of the sensed value of the biometric from the reference value.

4. A method as in claim 3, wherein the reference value is an ambient temperature.

5. A method as in claim 3, wherein detecting the stress level comprises computing a metric based on the deviations computed during successive time windows.

6. A method as in claim 5, wherein the metric is a function of a difference between deviations of two successive time windows.

7. A method as in claim 1, wherein detecting the stress level further comprises inferring a stress state transition based on a dynamic Bayesian Network model.

8. A method as in claim 7, wherein the dynamic Bayesian Network model includes a number of stress states each relating to a different one of a plurality of activities.

9. A method as in claim 7, wherein the stress state transition is inferred a conditional probability computed using the Chapman-Kolmogorov equation.

10. A method as in claim 1, wherein detecting the stress level further comprises inferring a stress state transition based on a time extension of a Bayesian Network (TBN) model.

11. A method as in claim 10, wherein the TBN model at any given time step comprises (i) a cause, corresponding to an activity, (ii) a hidden stress state; and (c) a plurality of observable values, each being a function of the sensed values of the biometric.

12. A method as in claim 10, wherein the TBN model is updated at the given time step by

(a) predicting a stress state for the given time step based on a conditional probability of the hidden stress state of an immediate previous time step given the observable values of one or more previous time steps, including the immediately previous time step; and

(b) computing the hidden stress state of the given time step using the predicted stress state of the given time step and the observable values of the given time step.

13. A method as in claim 11, wherein the TBN model includes a state transition model, an inference model and a detection model.

14. A method as in claim 13, wherein the state transition model is based on a conditional probability of the hidden stress state at the given time step, given the hidden stress state of an immediately previous time step.

15. A method as in claim 13, wherein the detection model is based on a conditional probability of the observable values of an immediately previous time step, given the hidden stress state of the immediately previous time step.

16. A method as in claim 13, wherein the inference model is based on a conditional probability of the hidden stress state of an immediately previous time step, given the observable values of one or more previous time steps, including the observable values of the immediately previous time step.

17. A method as in claim 11, wherein missing parameters of the TBN model is provided by a procedure based on the expectation maximization algorithm.

18. A method as in claim 1, wherein the method is implemented in an application software executable on a wireless handheld device.