Frame erasure pattern feedback for real-time perceptual quality estimation

The perceptual speech quality in modern communication systems is severely affected by "bad" received speech frames. The severity of quality degradation however varies from one speech frame to another depending on acoustic contents of each frame. In this paper, a scheme for real-time estimation of perceptual speech quality is proposed. In this scheme, the receiver sends the information back to the transmitter whether the previously received frame was "good" or "bad". This information is used in conjunction with perceptual evaluation of speech quality (PESQ) algorithm to estimate the receiver speech quality at the transmitter.