Nonlinear quantizer design in data conversion systems using the unscented transform
José Edil Guimarães de Medeiros, Sandro A. P. Haddad · 2017
This paper presents a novel approach to the design of signal specific quantizers based on the Unscented Transform. This is done in such a way that we preserve the statistical moments of the input signal on the quantization process. We define the design methodology for arbitrary signals and present the analysis of the arcsine distribution quantizer. For this case, we show that a 16dB improvement can be achieved in THD for quantizers in the range of 4 to 12 bits without degrading overall SNR.