Biological signals refer to measurable records of spatial and/or temporal events that happen inside a living system. Such signals could capture the electrical, chemical, or mechanical activities in an organism, and could thus be analyzed to derive valuable insights into physiological processes, behavioral patterns, and evolutionary relationships across diverse organisms. But analysis of biological signals can be challenging due to their noisy and high-dimensional nature. Our research focuses on addressing some of these challenges when analyzing two types of biological signals — genomic and electroencephalogram (EEG) signals — with the help of a careful combination of digital signal processing techniques with machine/deep learning models.
In the context of genomic signals, we would like to understand the genetic variations between different species at various levels of hierarchical/taxonomic organization. Existing genomic signal processing (GSP) based classification models can predict species from their genome sequences, but the biological interpretability of the signal-based features is not fully understood. This study aims to improve the accuracy of such GSP-based classification models and, importantly, to uncover biologically significant features that help improve the classification, ultimately shedding light on species diversity and genome evolution.
In the context of EEG signals, previous research has focused on identifying neural correlates of speech and music using signal processing and machine learning algorithms. However, there is a significant research gap when it comes to studying Indian languages and music. Our study aims to address this gap by decoding speech- and music-specific signatures from EEG signals, with a particular emphasis on Indian languages and music stimuli. By investigating the cognitive processing of speech and music in this context, we aim to enhance our understanding of the unique characteristics and neural mechanisms underlying language and music perception.