Non invasive human stress detection using key stroke dynamics and pattern variations

Stress is considered as a very harmful health problem in the modern world. A large proportion of sick leave in the industrial world is believed to be related to stress. While some level of stress can act as a positive factor, extensive exposure to high levels of stress can have detrimental effects on one´s health. Depression, panic attacks, high blood pressure, diabetes, heart problems are few of such diseases that can be initiated or worsened by stress. With the increasing people centricity in contemporary developments of computer science, Affective Computing has become a popular research area. According to the existing research, affective computing has shown positive results in detecting human stress. Stress detection has been tackled in various approaches including heart rate variability (HRV), skin conductance (SC), pupil diameter (PD) based detection, Finger Temperature (FT) and etc. Our focus in this study is to utilize a readily available yet underutilized resource in Affective Computing, key stroke dynamics (KSD). Recent developments in KSD based affective computing and Biometrics research proves that key stroke variations is a very powerful source of input that provides a valuable insight about an individual´s psychological and emotional states. Our methodology suggests a personalised approach in detecting stress levels through key stroke variations. An application specific Individual key stroke pattern profile is created for an individual based on his normal typing patterns. This profile consists of trained average values for a set of typing features. Real time stress specific deviations of these features are analysed in order to arrive at the individual stress level. The remainder of the paper is structured as follows. First we introduce the research problem that this study is attempting to address. Then the significance of our approach is brought to the attention. The related work section tries to analyse and evaluate the existing related literature r- volving around this research area. After that we present the details of our approach through the methodology section. The experiment section describes about the experiment conducted to gather the keystroke data that are to be analysed as stress and non-stress data samples. A brief overview of the results of the experiment is provided towards the end of the paper.