This project presents a parallel driver designed to address the challenges of respiratory rate (RR) estimation in real-world environments, combining multi-core architectures with parallel and high-performance techniques. Our proposed system employs a non-negative matrix factorization (NMF) approach to mitigate the impact of noise interference in the input signal. This NMF approach is guided by pre-trained bases of respiratory sounds and incorporates an orthogonal constraint to enhance accuracy. The proposed solution is tailored for real-time processing on low-power hardware, demonstrating promising outcomes in terms of accuracy and computational efficiency.
Our proposed solution builds upon the principles of NMF and parallel computing. The parallel driver efficiently handles the challenges posed by noisy environments, leveraging multi-core architectures and high-performance techniques. The NMF-based algorithm incorporates pre-trained bases of respiratory sounds and an orthogonal constraint to enhance accuracy. The implementation is based on a combination of C and Python and is optimized for real-time processing on low-power hardware, ensuring practical applicability in diverse healthcare settings.
To utilize our noise-tolerant NMF-based parallel algorithm for respiratory rate estimation, follow these steps:
- Clone the repository
- Refer to the following instructions to set up dependencies
- Execute the provided scripts for a comprehensive understanding of the library's usage
Before you proceed with the installation, ensure that you have the following prerequisites:
-
C Compiler: Install a C compiler on your system. If you haven't installed one yet, you can use the GNU Compiler Collection (GCC) for compatibility.
-
BLAS/LAPACK Library: Obtain and install a distribution of the BLAS/LAPACK library. This library is crucial for performing efficient numerical linear algebra operations.
-
FFTW Library: Download and install the FFTW library, which is essential for fast Fourier transform computations. Ensure proper configuration for optimal performance.
-
Python 3: Make sure you have Python 3 installed on your system. This is required for running the project's Python script.
Follow these steps to compile the C code:
-
Open your terminal using a bash-compatible shell.
-
Run the following commands:
make
Ensure that you properly link against the BLAS, LAPACK, and FFTW libraries, as well as the math library during the compilation process.
To run the project, execute the compiled binary along with the Python script. Follow these steps:
-
Open your terminal using a bash-compatible shell.
-
Run the following commands:
python RRmain.py path/ number_iterations
Replace path/ with the path to your input audio file and number_iterations with the desired number of iterations for the algorithm.
Ensure that the input audio files are in monochannel format and use a 16-bit per sample resolution for compatibility with the project.
We welcome contributions to enhance the capabilities of our noise-tolerant NMF-based parallel algorithm for respiratory rate estimation. Fork the repository, make improvements, and submit a pull request to the master branch.
If you use this algorithm in your research, please cite the following paper:
P. Revuelta-Sanz, A.J. Muñoz-Montoro, J. Torre-Cruz, F.J. Canadas-Quesada, J. Ranilla. "Noise-tolerant NMF-based parallel algorithm for respiratory rate estimation", 2023.
This project is licensed under the MIT License.
For inquiries or further information, please contact José Ranilla at [email protected].