Open AccessDynamic Quantization of Digital Filter Coefficients
Author(s) -
В. Н. Бугров
Publication year - 2021
Publication title -
trudy učebnyh zavedenij svâzi
Language(s) - English
Resource type - Journals
eISSN - 2712-8830
pISSN - 1813-324X
DOI - 10.31854/1813-324x-2021-7-2-8-17
Subject(s) - quantization (signal processing) , digital filter , mathematics , algorithm , filter design , control theory (sociology) , finite impulse response , transfer function , computer science , filter (signal processing) , artificial intelligence , engineering , control (management) , electrical engineering , computer vision
The possibility of quantizing the coefficients of a digital filter in the concept of dynamic mathematical
programming, as a dynamic process of step-by-step quantization of coefficients with their discrete optimization at
each step according to the objective function, common to the entire quantization process, is considered. Dynamic
quantization can significantly reduce the functional error when implementing the required characteristics of a lowbit digital filter in comparison with classical quantization. An algorithm is presented for step-by-step dynamic
quantization using integer nonlinear programming methods, taking into account the specified signal scaling and the
radius of the poles of the filter transfer function. The effectiveness of this approach is illustrated by dynamically
quantizing the coefficients of a cascaded high-order IIR bandpass filter with a minimum bit depth to represent integer
coefficients. A comparative analysis of functional quantization errors is carried out, as well as a test of the quantized
filter performance on test and real signals.