Abstract:
Voice over Internet Protocol, also known as VoIP allows conversion of analog audio signals into digital data that can be transmitted over the Internet. By doing this, the VoIP has the capability to use a standard Internet connection to make long distance calls worldwide at a cheaper rate. Delivering high quality voice communications over IP networks is a challenge because these networks have none of the characteristics that enable the PSTN to provide toll quality voice service. In this work a more effective and efficient method for voice communication over lossy packet networks by determining the achievable region for distortion in voice signals sent over the Internet via multi-path is proposed. A distortion measure in packet voice communication for MD-LVQ is presented. Distortion in voice communication is intolerable especially when it exceeds an acceptable region. The Internet is a best efforts network which is inhospitable to real-time communication due to its bursty nature. Link outages, congestion, node failures etc contribute to delay, jitter and packet loss on the Internet, these lead to distortion which degrade signals. To combat this unpredictable nature of the Internet and improve the quality of service, path diversity with multiple description coding was used. In this research a multi-path model for voice communication over lossy packet networks using lattice vector quantization was designed and a mathematical model for estimating packet distortion for multiple description voice signals was developed. A scenario where multiple independent paths are available between the sender and the receiver is considered. Retransmissions were not used when packets were lost in the network but multiple descriptions of the signals were sent over the multiple independent paths. The model developed shows that distortion in voice communication via the Internet can be reduced to an acceptable level by using lattice-based representation for multiple description coding with multi-path.
