Institute for Telecommunication Sciences / Research / Quality of Experience / Audio Quality Research / Examples of Digitally Compressed Speech

Examples of Digitally Compressed Speech

Speech Coder Speech Sample AD L(AD)
Original Speech
(No coding)		 source.wav 0.00 0.96
ITU Rec. G.711,
64 kbps µ-law PCM		 g711.wav 0.86 0.90
ITU Rec. G.726,
32 kbps µ-law ADPCM		 g726.wav 1.62 0.81
ITU Rec. G.728,
16 kbps LD-CELP		 g728.wav 1.82 0.77
GSM,
13 kbps RPE-LTP		 gsm.wav 1.66 0.80
2.4 kbps MELP melp.wav 3.09 0.31
US Federal Standard 1015,
2.4 kbps LPC10e		 lpc10e.wav 3.89 0.49

The third column of this table contains typical mean values of auditory distance (AD), as calculated by MNB Structure 2, defined in: S. Voran, " Objective Estimation of Perceived Speech Quality, Part I: Development of the Measuring Normalizing Block Technique" and "Part II: Evaluation of the Measuring Normalizing Block Technique," IEEE Transactions on Speech and Audio Processing, July 1999, or S. Voran, "Objective Estimation of Perceived Speech Quality Using Measuring Normalizing Blocks," NTIA Report 98-347, April 1998. The fourth column contains typical mean values of AD after transformation by a logistic function, as defined in those same documents.

Note that larger values of AD indicate that the speech is farther (in the perceptual sense) from the original speech. Larger values of L(AD) indicate higher estimated perceived speech quality.