Exploiting AMR-WB Audio Bandwidth Extension for Quality and Capacity Increase

Abstract

Audio bandwidth extension in AMR-WB to twice of that used in AMR-NB provides essential subjective speech quality improvements, while the link level performance in GERAN networks for codec modes of comparable source bit rate is similar. This study analyzes the effect of improved audio perception as well as the impact of channel errors and call drops on the network performance. Profound system level simulations for relaxed 4times3 and 3times3 frequency re-use as well as tight 1times1 reuse have been performed In 4times3 and 3times3 re-use networks, whose capacity is limited by hard-blocking, the audio advantage provided by AMR-WB is entirely transformed into quality improvements. These amount almost one third on the speech quality indicator (SQI) scale ranging from zero to one. The capacity of 1times1 re-use networks is primarily limited through soft-blocking criteria. Four different quality criteria have been applied: SQI, frame erasure rate (FER), bad quality probability (BQP), and call drop rate (CDR). If only the subjective speech quality criteria were taken into account an increase in network capacity from 21% Erlang fractional load (EFL) for AMR-NB to 32% for AMR-WB is feasible. Since this quality indicator is rather related to general speech quality impression than to intelligibility, additional FER based quality criteria and CDR have been applied. Requiring additionally BQP lower than 5% and CDR lower than 2% limits the capacity of AMR-WB tight re-use networks to 21% EFL. Exactly the same capacity is achieved by AMR-NB, for which subjective speech quality is the more restrictive criterion. Exchanging the BQP criterion by the more restrictive criterion of mean FER per call lower than 2% for 95% of the subscribers leads to capacity advantages for AMR-NB due to lack of sufficiently robust AMR-WB codec modes and the higher latency in codec mode adaptation using tandem free operation. Results indicate that tight re-use networks should not exploit first glance quality advantage of AMR-WB. Instead for the definition of admission control thresholds FER and BQP criteria shall be taken into consideration in addition to subjective speech quality impression.