Electroencephalogram (EEG) signal complexity quantifies the brain dynamic and yields many different features to diagnose many psychotic disorders. The aim of this work is to analyze EEG to detect schizophrenia in comparison with normal subjects using EEG signal complexity, at various conditions such as rest and mental activity. In order to stimulate mental activity, this work proposes a two modified odd ball paradigms. In this research work, 55 schizophrenia subjects and 23 normal subjects together 78 subjects are included. EEG is recorded under resting state with eyes closed and during the stimuli applied. Shannon entropy, Spectral entropy, Information entropy, Higuchi’s Fractal Dimension, Kolmogorov complexity and Approximate Entropies are considered as features and are analyzed in two aspects. One is at rest condition and the other is during application of stimulus for mental activity. The signal complexity is more for schizophrenia compared to the normal group during different mental states and it is more dominant during mental activity with p<0.0001 for the features Higuchi’s Fractal Dimension, Kolmogorov complexity and Approximate Entropy. The highest classification accuracy 88.5% is obtained when features of both stimulus are considered together. This work suggests that the EEG signal complexity during mental activity can be used to identify schizophrenia subjects.