This is an FPGA design for extracting four different envelograms from phonocardiogram (PCG) recordings. The envelograms are the Power Spectral Density (PSD) envelope, the Hilbert envelope, the Wavelet envelope and the Homomorphic envelogram.
More details can be found in this paper that uses such features for segmenting PCGs: Deep Convolutional Neural Networks for Heart Sound Segmentation - IEEE Journal of Biomedical and Health Informatics
Table of Contents
This project requires Vivado 2023.2 and Vitis Model Composer.
data_analysis/ folder contains python notebook used to analyze the IP operation emulations using recordings in audio samples/. ip_repo is the repository to be used to include the IP in a design within Vivado.
env_extQ16.slx/ and variables16.m are the Matlab/VMC design and the required variables, respectively.
All the results obtained from the tests are explained in the following paper: FPGA Design of Digital Circuits for Phonocardiogram Pre-Processing Enabling Real-Time and Low-Power AI Processing.
@INPROCEEDINGS{10741821,
author={Ragusa, Domenico and Rodriguez-Almeida, Antonio J. and Gazzoni, Marco and Torti, Emanuele and Marenzi, Elisa and Fabelo, Himar and Callico, Gustavo M. and Leporati, Francesco},
booktitle={2024 27th Euromicro Conference on Digital System Design (DSD)},
title={FPGA Design of Digital Circuits for Phonocardiogram Pre-Processing Enabling Real-Time and Low-Power AI Processing},
year={2024},
volume={},
number={},
pages={588-595},
keywords={Point of care;Signal processing algorithms;Transforms;Standardization;Real-time systems;Topology;Convolutional neural networks;Artificial intelligence;Phonocardiography;Digital circuits;Cardiovascular diseases;FPGA;Digital signal processing;Point-of-care;Medical device;Pre-processing;Phonocardiogram;Artificial intelligence},
doi={10.1109/DSD64264.2024.00084}
}