Neural Network-Based Post-Processing for the Enhancement of Transcoded Media
// Datensicherheit // Datenübertragung // Informations- und Kommunikationstechnik // Telecommunications
Ref-Nr: 16622
Einleitung / Abstract
The invention, by the application of neural networks, enables a quality of reception improvement of digital audio data.
Fig. 1: General flow chart of the postprocessor after transmission or storage of media.
Hintergrund
Transcodes audio and/or video signals always suffer from some quality loss due to the quantization/coding error that is measurable and oftentimes also perceivable. Postprocessing attempts to improve the quality of the transcodes signal after transmission or storage.
Problemstellung
So far, the existing post-processing methods are codec-specific or of inadequate quality. Therefore, there is a need for a post-processor that is effectivley suitable for a wide range of employed codecs or code combinations, and can be used on the receiver side without having to make adjustments to the encoder or to the decoder.Lösung
The invention uses a neural network as a post-processor for reconstructing enhanced audio and/or video signals. The method according to the invention is applicable to all types of codecs. The neural network is trained in advance for codecs of interest.Vorteile
Reduction of the quantization error
No changes to the encorder or to the decorder needed
Applicable to all networks and receivers
Use or neural networks
Anwendungsbereiche
Mobile communication, VoIP, music and video streaming, listening to MP3 / AAC, JPEG / MPEG decoderSie können dieses Fenster schließen. Ihre Suchergebnisse finden Sie in dem vorherigen Fenster
