David Sandberg

Multi-Rate Uplink Channel Prediction and Enhanced Link Adaptation for VoLTE

The paper presents an improvement of the long term evolution (LTE) link adaptation for voice over LTE (VoLTE). The underlying problem is associated with the speech frame structure of VoLTE, where the transmissions are typically scheduled every 40 ms, provided that discontinuous reception (DRX) is on. The channel estimates of each user are hence outdated each time a new uplink scheduling decision is made. The paper therefore suggests the application of channel prediction for slowly moving mobiles. A special challenge is then the fact that the channel measurements may be sampled by the rates 1 ms, 8 ms, 20 ms, 40 ms and 160 ms, in case of segmentation, retransmission, a normal speech frame, DRX and voice inactivity, respectively. A new optimal and adaptive channel prediction algorithm is developed to solve this multi-rate prediction problem. The algorithm combines a continuous time autoregressive (AR) model with the recursive least squares algorithm. An extensive system simulation study shows capacity improvements of 5-10% for Doppler frequencies below 10 Hz. The measured voice quality is also improved. No performance degradation appears in any of the simulated scenarios.