An Energy-Efficient Haar Wavelet Transform Architecture for Respiratory Signal Processing

摘要

This brief proposes an energy-efficient Haar wavelet transform for respiratory signal processing. We analyze and develop a Haar level 5 (Haar-5) transform architecture for separating the frequency bands of respiratory signals. The fixed-point Haar-5 transform herein proposed employs multi-level M = 1 , M = 2 , and M = 3 Haar transforms for the composition of five resolution levels. The architectures were described in VHDL and synthesized in hardware targeting a 65nm CMOS standard cell library. Our investigation results show that most area- and energy-efficient Haar architecture (i.e., version H-IV) employs one M = 1 block and two M = 2 blocks. Hardware synthesis results show that the H-IV architecture proposal saves 38.19% of circuit area and 38.26% of power dissipation and energy per operation, compared among the other architectures herein investigated. Our H-IV hardware architecture proposal shows more than 1150 times of energy reduction than state of the art.