Amélioration de la robustesse de décodeurs de parole basés sur le modèle CELP en utilisant les informations retardées application au standard G.729 pour la voix sur IP
Voice over Internet applications is the new trend in telecommunications and networking industry today. Packetizing data/voice is done using the Internet protocol (IP). Various codecs exist to convert the raw voice data into packets. The coded and packetized speech is transmitted over the Internet. At the receiving end some packets are either lost, damaged or arrive late. This is due to constraints such as network delay (fitter), network congestion and network errors. These constraints degrade the quality of speech. Since voice transmission is in real-time, the receiver can not request the retransmission of lost or damaged packets as this will cause more delay. Instead, concealment methods are applied either at the transmitter side (coder-based) or at the receiver side (decoder-based) to replace these lost or late-arrival packets. This work attempts to implement a novel method for improving the recovery time of concealed speech The method has already been integrated in a wideband speech coder (AMR-WB) and significantly improved the quality of speech in the presence of jitter in the arrival time of speech frames at the decoder. In this work, the same method will be integrated in a narrowband speech coder (ITU-T G.729) that is widely used in VoIP applications. The ITUT G.729 coder defines the standards for coding and decoding speech at 8 kb/s using Conjugate Structure Algebraic Code-Excited Linear Prediction (CS-CELP) Algorithm.
- Génie – Mémoires