New Network and ATM Adaptation Layers for Real-Time Multimedia Applications: A Performance Study Based on Psychophysics

We present in this paper Network and ATM Adaptation Layers for real-time multimedia applications. These layers provide a robust transmission by applying per-cell sequence numbering combined with a selective Forward Error Correction (FEC) mechanism based on Burst Erasure codes. We compare their performance against a transmission over AAL5 by simulating the transport of an MPEG-2 sequence over an ATM network. Performance is measured in terms of Cell Loss Ratio (CLR) and user perceived quality. The proposed layers achieve an improvement on the cell loss figures obtained for AAL5 of about one order of magnitude under the same traffic conditions. To evaluate the impact of cell losses at the application level, we apply a perceptual quality measure to the decoded MPEG-2 sequences. From a perceptual point of view, the proposed AAL achieves a graceful quality degradation compared to AAL5 which shows a critical CLR value beyond which quality drops very fast. The application of a selective FEC achieves an even smoother image quality degradation with a small overhead.