Human beings can enumerate less than 5-6 items very quickly, confidently and accurately, even if the display is very short. If more items are presented we tend to guess and our answers are more tentative, slow and inaccurate. During my Ph.D. with Dr. S. Bapiraju at Center for Neural and Cognitive Sciences, University of Hyderabad, India, I developed a neuro-computational model for networks in lateral intra-parietal (LIP) region that can account for human enumeration performances across both the small and large ranges of numerosity as well as make some novel testable predictions (Sengupta et al,, 2014, Brain Res.). The model received support from empirical fMRI studies on enumeration and working memory looking at LIP region (Knops, Piazza, Sengupta, Eger, \& Melcher, 2014, J. Neurosci.). In subsequent behavioral work, we validated the computational model by testing the predictions (Sengupta et al., Attention, Perception, \& Psychophysics, 2017).  I have also developed a computational framework that can account for individuation (required for enumeration and general visual attention task involving multiple objects, Verma and Sengupta, Scientific Reports, 2023) and ensemble processing (as seen in perceptual averaging tasks (Sengupta, Springer LNCS, 2024, In Press).

Objects attended to in visual field are more easily sustained in working memory. Also, spatial attention and visual short-term memory share brain networks. Working memory is now established as a limited capacity system. For visual working memory (VWM), visual objects within scope of attention are stored for further processing or recall. However, it remains unclear whether all the features are stored with equal fidelity within the memory representation remains to be investigated more thoroughly. Our recent works (Sengupta et al, Visual Cognition, 2024, In Revision) we have shown interference effects in VWM fail to account for results where multiple features are probed for the same object representation maintained in working memory. This opens up new avenues of research for working memory mechanisms involving forgetting.  We have also modeled serial working memory and corresponding primacy and recency effects using a RNN (Sengupta et., al, Springer LNCS, 2024)

It has been seen that in an oddball paradigm that the oddball stimulus of the same duration as the stream of standard stimuli, is generally perceived as longer (Tse et al, 2004). The suggested mechanisms are based on internal clocks or attention based information accrual. I proposed a recurrent neural network model optimized for winner-take-all paradigm in order to simulate the oddball judgment. Model was able to account for both the subjective expansion of time for standard stimuli longer than 120 ms, as well as subjective contraction of time for standard stimuli smaller than 120 ms in duration. I tested the model predictions with different standard durations in a psychophysical experiment (Sengupta et. al., IEEE XPlore, In Press, 2024). I have also used the behavioral paradigm for EEG-based Brain-computer Interface (EEG-BCI) experiments (Sengupta, et al, IEEE XPlore, In Press, 2024). The philosophical implications of such a model has been explored in my previous work (Sengupta, 2018, JICPR). 

Our lab also develops experimental paradigms and algorithms (Janapati, Dalal, and Sengupta et. al.,  NanoLIFE, 2022) for various Brain-Computer interface technologies (Sengupta et. al., IEEE XPlore, 2024, In Press). More recently, we have also developed a hybrid EEG-Eye tracking based BCI system prototype for hands-free web browser (Sengupta et al., CRC Press, 2024, In Press).