There is not much information on the specifics of your application possibility.  So, it is not possible to provide specific help.  But, in a general response, my experience indicates that if a human can control something, putting those same human rules in FLC will automate the control with equivalent outcomes; but, with better and continual attention than humans. 

Often a first development of FLC embodies a superficial understanding, and the human should periodically observe the FLC performance and fine-tune, adjust, or add rules to better provide a comprehensive controller. 

I now understand that it is not control that you are doing it is modeling.  You want to use fuzzy logic, FL, to be able to take a person's recent experiences/situation and predict their level of stress. 

Your input variables are the separate categories of experience or situation or environment that might influence a person’s level of stress.  These factors might be “recreation”, “sleep”, “work”, “physical fitness”, “personal debt”, “number of dependents”, “support from a significant other”, etc.  But it is not a yes or no that matters for the factors, it is a quantification of aspects such as level, quality, and time since.     

First, use text and human logic to describe how the modeling relation might be done.  For example it is not the fact that a person got some sleep in the past day, but it is the amount of sleep (3 hours, 9 hours), the number of times for the sleep duration (1 sleep through, 2 sleep sections, all in 6 short catnaps) and the time duration since waking (2 hours, 18 hours) that would be the things that you would consider that influence the internal level of stress.  Inputs to the FL model would be amount of sleep in the past day (hours), number of sleeps, and time since last sleep.  These are all continuous valued or integer variables that quantify the quality/quantity/magnitude/persistence of the influence.  And, it is not the person's body weight, but the weight over or under (5 kg, 100 kg) a healthy weight for the person's height and type.  List these quantifiable attributes/influences.

One option to FL modeling would have you decide how to classify the numerical variables.  For example, any sleep duration between 7 and 9 hours might be categorized as “normal”, 5 to 7 “little”, less than 5 “very little”, 9 to 11 “excessive”, and over 11 “very excessive”.  If in 1 period “normal”, 2 to 3 “interrupted” and over 3 “poor”.  These define the memberships in each factor category (linguistic variable).  Calculate a membership value for each variable in each category.

Do the same for stress level.  Decide quantifications “low”, “normal”, “moderately high”, “high”, “very high”.  For each data point assign a membership of stress to each category.

Then use nonlinear regression to create a model that uses the input memberships to predict the stress membership. 

Thanks , for the helpful answer.

so, now fuzzy logic will be used to model stress levels , and nonlinear regression for the context information (the factors that affect stress levels)

In a computer any variable is discrete, even if it represents real-life continuous (in other words analog) signals or parameters. So, if you implement Fuzzy logic in a computer automatically you have to work with discrete values. The question is how discrete, or how many discrete values represent the variation of a parameter in a given domain. It may go from just two values (a Boolean variable) to a very high granularity where the discrete variable models very well the real-world continuous parameter.

As a conclusion, yes you can use Fuzzy logic for discrete parameters if the granularity (number of discrete values) of the discrete parameter is enough to model your problem.