Symbolic analysis of quantization errors in transform and subband coders
Frank Heinle, J. Lerzer, G. Runze · 2002
In general, transform and subband coders are designed so as to achieve perfect reconstruction in the ideal case. Nevertheless, in a real system there are many disturbing effects such as quantization in digital filters, intermediate coding stages, and noisy transmission channels which prevent perfect reconstruction. Usually, these effects are either neglected or modelled in a very simple way. In contrast, we present a new software tool which permits an automatic symbolic analysis of implemented filter banks using refined models of wordlength effects. Furthermore, we validate the symbolic outcomes by measurement results.