Signal and Noise Generation
ANF Generator — Arbitrary Noise Fields
Generates multi-channel noise signals with a predefined spatial coherence function. Supports spherically isotropic, cylindrically isotropic, and Corcos (wind-noise) coherence models. The mixing matrix is obtained by Cholesky or eigenvalue decomposition; three post-processing methods (smooth, balanced, balanced+smooth) based on the unitary Procrustes solution improve spectral smoothness and mix balance. Suitable for generating babble speech, factory noise, and wind noise in multi-sensor configurations.
The Python implementation is available here and can be installed via pip install anf-generator. The MATLAB implementation is available here.
- D. Mirabilii, S.J. Schlecht, E.A.P. Habets, Generating coherence-constrained multisensor signals using balanced mixing and spectrally smooth filters, Journal of the Acoustical Society of America, 149:1425, 2021.
INF Generator — Isotropic Noise Fields
Generates sensor signals for an arbitrary one- or three-dimensional array that result from a spherically or cylindrically isotropic noise field. Implements the algorithms described in Habets and Gannot (2007) and the associated internal report (2010).
The MATLAB implementation is available here.
- E.A.P. Habets and S. Gannot, Generating sensor signals in isotropic noise fields, Journal of the Acoustical Society of America, 122(6):3464–3470, 2007.
Wind Noise Generator
Generates multi-channel artificial wind noise signals. The MATLAB implementation models the complex spatial coherence of wind noise using the Corcos model, which depends on inter-microphone distance, airstream direction, and free-field flow velocity. The Python implementation extends this with wind speed profile-dependent characteristics, making it particularly suitable for generating training data for deep learning-based wind noise reduction systems.
The Python implementation is available here. The MATLAB implementation is available here.
- D. Mirabilii and E.A.P. Habets, Simulating multi-channel wind noise based on the Corcos model, Proc. International Workshop on Acoustic Signal Enhancement (IWAENC), 2018.
Anechoic Interferer Dataset Generator
A Python utility for generating mixtures of random, anechoic, non-stationary noise signals. Intended for use as interferer signals in speech enhancement and noise suppression experiments. The accompanying dataset as described in [4] is availbale on Zenodo.
The Python implementation is available here.
- P. Goetz, C. Tuna, A. Walther and E.A.P. Habets, AID: Open-source anechoic interferer dataset, Proc. International Workshop on Acoustic Signal Enhancement (IWAENC), 2022.