Spherical Microphone Array Processing Toolbox

Summary

• This repository contains a library for some of the features commonly used in spherical array microphone processing.
• These features are mostly used for direction-of-arrival (DOA) and six-degrees of freedom (6DoF) problems
• It is easy to add new features, datasets, microphones.
• This repository consists of two main interfaces: features as library and main script as feature extractor using config files.
• main.py is for extracting batch features from emulations of selected anechoic/music files on SMIR dataset over different positions and rooms.
• Batch extraction can easily be done for readily-prepared classes for datasets and microphones using integrated configuration system via hydra.

Supported

Features

• Spherical harmonic decomposition (SHD) (Rafaely, 2015)
• Direct path dominance (DPD) (Nadiri and Rafaely, 2014)
• Generalised Pareto distribution (GPD) fit over singular values from DPD (Olgun and Hacihabiboglu, 2019)
• Sparse plane wave decomposition (PWD) via orthogonal matching pursuit (OMP) using dictionary of Legendre kernels over HEALPix (Coteli and Hacihabiboglu, 2021)
• Residual energy test (RENT) (Coteli and Hacihabiboglu, 2021)

SMIR datasets

• METU SPARG AIR (Zenodo)
• BBC Maida Vale Impulse Response Dataset (Zenodo)
• TODO: TAU-SRIR DB (Zenodo)

Microphones

• mh Acoustics Eigenmike em32
• TODO: Zylia ZM-1

How to use

python main.py --help

Tasks

• Add support for real SMA recordings
• Fix fimin/fimax passing everywhere and discard empty npy portions
• Build a pipeline system
• Implement analyse functions
• wandb integration for analyse?
• Integrate room simulations

References

• B. Rafaely, Fundamentals of Spherical Array Processing, vol. 8. Berlin, Heidelberg: Springer Berlin Heidelberg, 2015. doi: 10.1007/978-3-662-45664-4.
• O. Nadiri and B. Rafaely, "Localization of Multiple Speakers under High Reverberation using a Spherical Microphone Array and the Direct-Path Dominance Test," in IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 22, no. 10, pp. 1494-1505, Oct. 2014, doi: 10.1109/TASLP.2014.2337846.
• O. Olgun, H. Hacihabiboglu, "Data-driven Threshold Selection for Direct Path Dominance Test, Proceedings of the 23rd International Congress on Acoustics, 2019, pp. 3313–3320.
• M. B. Coteli and H. Hacihabiboglu, “Sparse Representations With Legendre Kernels for DOA Estimation and Acoustic Source Separation,” IEEE/ACM Trans. Audio Speech Lang. Process., vol. 29, pp. 2296–2309, 2021, doi: 10.1109/TASLP.2021.3091845.