使用於即時心率估計系統之記憶體式FFT設計

Abstract

本文論提出一個用於穿戴式紡織電極的即時心率估計系統的記憶體式快速傅利葉轉換，心電圖訊號容易受到電極與身體間接觸鬆弛和走動的影響。也就是基線電位干擾。我們使用子空間追蹤技術消除基線電位的干擾，接加入絕對值運算來符合後面心率估計系統所需。心率估計系統使用自我相關度的方法來找出R-R間距並計算出心率。為了能達到即時的需求，我們使用藉由Power Method而得秩為一的適應性子空間追蹤技術。在心率估計方面，我們用FFT跟IFFT來實現自我相關度的計算。 在硬體實現方面，我們使用FPGA來做訊號處理與心率估測的部份。在FFT部份，由於心率估計系統是一個工作頻率不快的系統，因此我們採用記憶體式的架構來減少面積和功率的消耗。設計的結果經過編譯並下載至Altera EP2C35F672 Cyclone II FPGA 裝置上驗證

This thesis presents a memory-based fast Fourier transform (FFT) for a real time heart rate estimation system for wearable textile sensors. The electrocardiogram (ECG) is seriously affected by loose contact between biopotential electrodes and skin, and body movement. Also, it has baseline wander interference. For the baseline wander interference, we use the subspace approach to remove the baseline wander component, then uses the absolute operation for heart rate estimate. The method of heart rate estimation applies the correlation technique to find the R-R interval and estimates the heart rate. For the requirement of real time, we use rank one adaptive algorithm to find dominant eigenvalue by the power method. For the heart rate estimator, we use FFT and IFFT to achieve the correlation technique. In the hardware implementation, we use FPGA implementation to achieve the signal processing and heart rate estimation. In the proposed FFT, we adopt memory-based architecture to reduce area and power dissipation since heart rate estimation system is a low frequency system. The resulting design are compiled and download to the Altera EP2C35F672 Cyclone II FPGA device for real-time verification.