Packet loss is a major cause of voice quality degradation in VoIP transmissions with serious impact on intelligibility and user experience. This paper describes a system based on a generative adversarial approach, which aims to repair the lost fragments during the transmission of audio streams. Inspired by the powerful image-to-image translation capability of Generative Adversarial Networks (GANs), we propose bin2bin, an improved pix2pix framework to achieve the translation task from magnitude spectrograms of audio frames with lost packets, to non-corrupted speech spectrograms. In order to better maintain the structural information after spectrogram translation, this paper introduces the combination of two STFT-based loss functions, mixed with the traditional GAN objective. Furthermore, we employ a modified PatchGAN structure as discriminator and we lower the concealment time by a proper initialization of the phase reconstruction algorithm. Experimental results show that the proposed method has obvious advantages when compared with the current state-of-the-art methods, as it can better handle both high packet loss rates and large gaps. We make our code publicly available at: github.com/aircarlo/bin2bin-GAN-PLC.
A Time-Frequency Generative Adversarial based method for Audio Packet Loss Concealment / Aironi, C.; Cornell, S.; Serafini, L.; Squartini, S.. - (2023), pp. 121-125. (Intervento presentato al convegno 31st European Signal Processing Conference, EUSIPCO 2023 tenutosi a Helsinki, Finland nel 04-08 September 2023) [10.23919/EUSIPCO58844.2023.10290027].
A Time-Frequency Generative Adversarial based method for Audio Packet Loss Concealment
Aironi C.;Cornell S.;Serafini L.;Squartini S.
2023-01-01
Abstract
Packet loss is a major cause of voice quality degradation in VoIP transmissions with serious impact on intelligibility and user experience. This paper describes a system based on a generative adversarial approach, which aims to repair the lost fragments during the transmission of audio streams. Inspired by the powerful image-to-image translation capability of Generative Adversarial Networks (GANs), we propose bin2bin, an improved pix2pix framework to achieve the translation task from magnitude spectrograms of audio frames with lost packets, to non-corrupted speech spectrograms. In order to better maintain the structural information after spectrogram translation, this paper introduces the combination of two STFT-based loss functions, mixed with the traditional GAN objective. Furthermore, we employ a modified PatchGAN structure as discriminator and we lower the concealment time by a proper initialization of the phase reconstruction algorithm. Experimental results show that the proposed method has obvious advantages when compared with the current state-of-the-art methods, as it can better handle both high packet loss rates and large gaps. We make our code publicly available at: github.com/aircarlo/bin2bin-GAN-PLC.File | Dimensione | Formato | |
---|---|---|---|
A_Time-Frequency_Generative_Adversarial_Based_Method_for_Audio_Packet_Loss_Concealment.pdf
Solo gestori archivio
Tipologia:
Versione editoriale (versione pubblicata con il layout dell'editore)
Licenza d'uso:
Tutti i diritti riservati
Dimensione
482.02 kB
Formato
Adobe PDF
|
482.02 kB | Adobe PDF | Visualizza/Apri Richiedi una copia |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.