In order to be able to transmit real-time oriented voice applications via data packet networks, the data to be transmitted must be compressed. For voice compression, the International Telecommunication Union( ITU) has adopted standards for coding methods that offer different quality in speech intelligibility depending on the usable bandwidth. These compression methods are part of the H.323 protocol family, which includes several speech codecs.
As an evaluation criterion for the quality of the coded speech, the ITU has defined a measurement variable that is calculated from the average evaluation of different speech patterns by several groups of people: This evaluation variable is called Mean Opinion Score( MOS). The MOS scale ranges from 0 to 5, with the upper value representing good speech intelligibility. Depending on the coding method used, the effective bandwidth used for speech transmission can be reduced by up to 90%. It should be noted that all methods require a relatively extensive overhead for transmission. This overhead can exceed the data portion for speech several times over.
The ITU-T standards are the G recommendations G.711 to G.729, which use a wide variety of coding methods, bandwidths, delay times and MOS values, Mean Opinion Score (MOS). Of particular interest is the G.722 standard, which is used for quality-enhanced telephony and is deployed in Adaptive Multirate Wideband( AMR- WB).
Methods for voice compression include: Adaptive Delta Pulse Code Modulation( ADPCM), Code Excited Linear Prediction( CELP), GSM 06.10, Multiple Maximum Likehood Quantization (MPMLQ), Harmonic Vector Excitation( HVXC) and Linear Predictive Coding ( LPC).