Graph-Constrained Sequence Decoding (GCSD) is an ECG-only framework for robust inter-beat interval (IBI) estimation under motion artifact. The framework consists of three main stages: candidate peak generation, graph-based sequence modeling, and rhythm-aware path optimization.
GCSD treats outputs from multiple ECG detectors as weak evidence rather than final decisions. Candidate R-peaks are merged into an over-complete set, organized as nodes in a directed acyclic graph, and decoded with rhythm-aware optimization to recover a physiologically consistent beat sequence for R-peak, IBI, and heart-rate estimation.
The GCSD paper is currently under review in ACM Transactions on Computing for Healthcare (ACM Health).
We will release a Python package for GCSD. After release, it will be installable with:
pip install gcsd-ecg
Shovito Barua Soumma
Graduate Research Associate
PhD Candidate
Currently I am working on building and optimizing deep learning models for wearable sensors data.