DNN-based mask estimation for distributed speech enhancement in spatially unconstrained microphone arrays

Deep neural network (DNN)-based speech enhancement algorithms in microphone arrays have now proven to be efficient solutions to speech understanding and speech recognition in noisy environments. However, in the context of ad-hoc microphone arrays, many challenges remain and raise the need for distributed processing. In this paper, we propose to extend a previously introduced distributed DNN-based time-frequency mask estimation scheme that can efficiently use spatial information in form of so-called compressed signals which are pre-filtered target estimations. We study the performance of this algorithm named Tango under realistic acoustic conditions and investigate practical aspects of its optimal application. We show that the nodes in the microphone array cooperate by taking profit of their spatial coverage in the room. We also propose to use the compressed signals not only to convey the target estimation but also the noise estimation in order to exploit the acoustic diversity recorded throughout the microphone array.

Domaines

Traitement du signal et de l'image [eess.SP]

Fichier principal

furnon.pdf (3.57 Mo)

bi_vs_wi.png (58.21 Ko)

crossconfig_living_sn2.png (88.13 Ko)

crossconfig_meeting_sn2.png (58.33 Ko)

crossconfig_random_sn2.png (58.25 Ko)

fs_vs_fsdn_random2.png (161.84 Ko)

living_room.png (47.63 Ko)

mask_on_z_distant.pdf (17.22 Ko)

mask_on_z_local.pdf (17.19 Ko)

meeting_room.png (57.43 Ko)

mn_random_best_in2.png (88.78 Ko)

mn_random_best_out2.png (65.68 Ko)

mn_random_worst_in2.png (65.8 Ko)

noise_fs2.png (74.8 Ko)

noise_ssn2.png (74.64 Ko)

perf_analysis_rt_dstoi.png (100.36 Ko)

perf_analysis_rt_stoi.png (123.83 Ko)

perf_analysis_sir_dsir.png (131.71 Ko)

perf_analysis_sir_sir.png (138.89 Ko)

random_room.png (62.57 Ko)

sigex.pdf (292.45 Ko)

sn_vs_mn_living2.png (75.54 Ko)

sn_vs_mn_meeting2.png (58.67 Ko)

sn_vs_mn_random2.png (58.31 Ko)

specs_input.png (412.73 Ko)

specs_irm.png (410.12 Ko)

specs_m1.png (301.81 Ko)

specs_m2.png (242.5 Ko)

specs_output.png (365.34 Ko)

specs_z.png (384.68 Ko)

step2.pdf (31.75 Ko)

tango_nn_v2.pdf (46.83 Ko)

Origine	Fichiers produits par l'(les) auteur(s)

Origine	Fichiers produits par l'(les) auteur(s)

Origine	Fichiers produits par l'(les) auteur(s)

Nicolas Furnon : Connectez-vous pour contacter le contributeur

https://hal.science/hal-02985867

Soumis le : vendredi 20 août 2021-08:58:21

Dernière modification le : mercredi 20 novembre 2024-15:28:17

Dates et versions

hal-02985867 , version 1 (02-11-2020)

hal-02985867 , version 2 (19-05-2021)

hal-02985867 , version 3 (20-08-2021)

Identifiants

HAL Id : hal-02985867 , version 3
ARXIV : 2011.01714
DOI : 10.1109/TASLP.2021.3092838

Citer

Nicolas Furnon, Romain Serizel, Slim Essid, Irina Illina. DNN-based mask estimation for distributed speech enhancement in spatially unconstrained microphone arrays. IEEE/ACM Transactions on Audio, Speech and Language Processing, 2021, 29, pp.2310 - 2323. ⟨10.1109/TASLP.2021.3092838⟩. ⟨hal-02985867v3⟩

Exporter

BibTeX XML-TEI Dublin Core DC Terms EndNote DataCite

Collections

CNRS INRIA GRID5000 UNIV-LORRAINE INRIA2 LORIA LORIA-NLPKD LTCI IDS S2A IP_PARIS SILECS ANR INSTITUT-MINES-TELECOM

316 Consultations

414 Téléchargements