Analog VLSI Processor Implementing the Continuous Wavelet Transform
Abstract
We present an integrated analog processor for real-time wavelet decom(cid:173) position and reconstruction of continuous temporal signals covering the audio frequency range. The processor performs complex harmonic modu(cid:173) lation and Gaussian lowpass filtering in 16 parallel channels, each clocked at a different rate, producing a multiresolution mapping on a logarithmic frequency scale. Our implementation uses mixed-mode analog and dig(cid:173) ital circuits, oversampling techniques, and switched-capacitor filters to achieve a wide linear dynamic range while maintaining compact circuit size and low power consumption. We include experimental results on the processor and characterize its components separately from measurements on a single-channel test chip.
Cite
Text
Edwards and Cauwenberghs. "Analog VLSI Processor Implementing the Continuous Wavelet Transform." Neural Information Processing Systems, 1995.Markdown
[Edwards and Cauwenberghs. "Analog VLSI Processor Implementing the Continuous Wavelet Transform." Neural Information Processing Systems, 1995.](https://mlanthology.org/neurips/1995/edwards1995neurips-analog/)BibTeX
@inproceedings{edwards1995neurips-analog,
title = {{Analog VLSI Processor Implementing the Continuous Wavelet Transform}},
author = {Edwards, R. Timothy and Cauwenberghs, Gert},
booktitle = {Neural Information Processing Systems},
year = {1995},
pages = {692-698},
url = {https://mlanthology.org/neurips/1995/edwards1995neurips-analog/}
}