<p>Anxiety is a mental health condition associated with negative affect. Chronic anxiety leads to anxiety disorder causing significant impairment in important areas of functioning and reducing quality of life. With the growth of computer vision technology, eye movement features have become popular in the diagnosis affective conditions like anxiety because of their non-invasiveness. Our objective is to investigate the blink patterns to determine the presence of anxiety through emotional elicitation. Blink data of 44 participants from different regions of the country within the age group of 18–30&#xa0;years were collected using an experimental set up to elicit major emotions, i.e., joy, sad, disgust, fear, anger and to record participant’s visual response. Both statistical analysis as well as binary classification using machine learning technique were carried out on the blink data to find presence of anxiety. Subjecting the blink data to statistical tests revealed that blink rates of normal and anxiety affected groups differed significantly during sad and disgust emotional states when considered separately as well as when considered together. Applying machine learning techniques namely Bayes’ Net, k-NN classifier, SVM, logistic regression and decision tree on blink data resulted in detection accuracies ranging from 88 to 94% using tenfold cross validation. These findings suggest that prevalence of anxiety problems can be detected during onset stage by making use of non-invasive features like blink patterns and hence may be exploited extensively to be considered as biomarkers. The availability of quality data in terms of minimal noise, high resolution, large sample size, wide demographic characteristics possesses a key challenge. Machine learning approach driven by computer vision technology provides ample scope to explore all such eye movement features that can prevent further progress of similar mental health problems.</p>