Abstract
The growth of computer networks has led to the development of the Packet Switching and Packet Broadcast Techniques. Especially useful for bursty traffic, they are of use for speech which is inherently so, with about 50% of the time occupied by silence periods. In Packet Transmission of speech, we digitize the speech and transmit packets of fixed or variable size and reassemble the bits at the destination.
Unlike transmission of data, where the bit error rate is most important, in speech the prime consideration is delay. Depending upon the digitizing method, we have different error tolerance limits.
In this paper, we first address the problems involved in Speech Transmission in packets, and then examine a few specific cases of Packet Voice Transmission. The concepts of TASI are also inherent in Packet Voice Transmission, where silence periods are eliminated. This involves the use of efficient speech activity detectors. Since there is a certain amount of error tolerance present in digitized speech (albeit in varying degrees), this aspect can be used while reassembling to reduce delay. In our results, we present the performance of a few systems under different conditions of packet loss, delay etc; the prime consideration used in comparison being the subjective quality.
Additional information
Notes on contributors
C.V. Chakravarthy
Chakravarthy, C V (Dr). Obtained the BE and ME degrees from the Indian Institute of Science in 1970 and 1972 respectively and the PhD degree from the Indian Institute of Technology (Kharagpur) in 1977. He was a lecturer in the E & ECE department of IIT, Kharagpur from 1975 and at present is an assistant professor in the ECE Department of the Indian Institute of Science. His interests are in Speech Coding, Packet Communication and application of Microprocessors to Communication systems.