GEV Beamforming Supported by DOA-Based Masks Generated on Pairs of Microphones

Grondin, François; Lauzon, Jean-Samuel; Vincent, Jonathan; Michaud, François

doi:10.21437/interspeech.2020-2687

“…In future work, additional functionalities will be added to ODAS, including new algorithms that rely on deep learning based methods, as machine learning has became a powerful tool when combined with digital signal processing for sound source localization Chakrabarty and Habets (2019) , speech enhancement Valin (2018) ; Valin et al (2020) and sound source classification Ford et al (2019) ; Grondin et al (2019) . Additional beamforming methods could also be implemented, including Minimum Variance Distortionless Response (MVDR) beamformer Habets et al (2009) , and generalized eigenvalue (GEV) beamforming Heymann et al (2015) ; Grondin et al (2020) , as these approaches are particularly suited for preprocessing before automatic speech recognition. ODAS would also benefit from ego-noise suppression algorithms to mitigate the impact of motor noise while doing sound source localization, tracking and separation.…”

Section: Discussionmentioning

confidence: 99%

ODAS: Open embeddeD Audition System

Grondin

¹

,

Létourneau

²

,

Godin

³

et al. 2022

Self Cite

View full text Add to dashboard Cite

Artificial audition aims at providing hearing capabilities to machines, computers and robots. Existing frameworks in robot audition offer interesting sound source localization, tracking and separation performance, although involve a significant amount of computations that limit their use on robots with embedded computing capabilities. This paper presents ODAS, the Open embeddeD Audition System framework, which includes strategies to reduce the computational load and perform robot audition tasks on low-cost embedded computing systems. It presents key features of ODAS, along with cases illustrating its uses in different robots and artificial audition applications.

show abstract

“…In future work, additional functionalities will be added to ODAS, including new algorithms that rely on deep learning based methods, as machine learning has became a powerful tool when combined with digital signal processing for sound source localization [33], speech enhancement [34], [35] and sound source classification [36], [37]. Additional beamforming methods could also be implemented, including Minimum Variance Distortionless Response (MVDR) beamformer [38], and generalized eigenvalue (GEV) beamforming [39], [40], as these approaches are particularly suited for preprocessing before automatic speech recognition.…”

Section: Discussionmentioning

confidence: 99%

ODAS: Open embeddeD Audition System

Grondin¹,

Létourneau²,

Godin³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Artificial audition aims at providing hearing capabilities to machines, computers and robots. Existing frameworks in robot audition offer interesting sound source localization, tracking and separation performance, but involve a significant amount of computations that limit their use on robots with embedded computing capabilities. This paper presents ODAS, the Open embeddeD Audition System framework, which includes strategies to reduce the computational load and perform robot audition tasks on low-cost embedded computing systems. It presents key features of ODAS, along with cases illustrating its uses in different robots and artificial audition applications. I. INTRODUCTIONSimilarly to artificial/computer vision, artificial/computer audition can be defined as the ability to provide hearing capabilities to machines, computers and robots. Vocal assistants on smart phones and smart speakers are now common, providing a vocal interface between people and devices [1]. But as for artificial vision, there are still many problems to resolve for endowing robots with adequate hearing capabilities, such as ego and non-stationary noise cancellation, mobile and distant speech and sound understanding [2]- [6].Open source software frameworks, such as OpenCV [7] for vision and ROS [8] for robotics, greatly contribute in making these research fields evolve and progress, allowing the research community to share and mutually benefit from collective efforts. In artificial audition, two main frameworks exist: • HARK (Honda Research Institute Japan Audition for Robots with Kyoto University 1 ) provides multiple modules for sound source localization and separation [9]-[11]. This framework is mostly built over the FlowDesigner software [12], and can also be interfaced with speech recognition tools such as Julius [13] and Kaldi [14], [15]. HARK implements sound source localization in 2-D using variants of the Multiple Signal Classification (MUSIC) algorithm [16]-[18]. HARK also performs geometrically-constrained higher-order decorrelation-based source separation with adaptive *This work was supported by FRQNT -Fonds recherche Québec Nature et Technologie.

show abstract

“…We denote the SCMs for the target speech, the interfering noise and the resulting mixture as R XX [k] ∈ C M ×M , respectively, where M stands for the number of microphones. The SCMs for speech and noise can be obtained using time-frequency masks [56,103,104].…”

Section: A31 Multi-microphone Signal Processingmentioning

confidence: 99%

SpeechBrain: A General-Purpose Speech Toolkit

Ravanelli,

Parcollet,

Plantinga

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

SpeechBrain is an open-source and all-in-one speech toolkit. It is designed to facilitate the research and development of neural speech processing technologies by being simple, flexible, user-friendly, and well-documented. This paper describes the core architecture designed to support several tasks of common interest, allowing users to naturally conceive, compare and share novel speech processing pipelines. SpeechBrain achieves competitive or state-of-the-art performance in a wide range of speech benchmarks. It also provides training recipes, pretrained models, and inference scripts for popular speech datasets, as well as tutorials which allow anyone with basic Python proficiency to familiarize themselves with speech technologies.

show abstract

GEV Beamforming Supported by DOA-Based Masks Generated on Pairs of Microphones

Cited by 8 publications

References 0 publications

ODAS: Open embeddeD Audition System

ODAS: Open embeddeD Audition System

ODAS: Open embeddeD Audition System

SpeechBrain: A General-Purpose Speech Toolkit

Contact Info

Product

Resources

About