13.07.2015 Views

Thesis - Instituto de Telecomunicações

Thesis - Instituto de Telecomunicações

Thesis - Instituto de Telecomunicações

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

88 CHAPTER 4. SIGNAL PROCESSING AND FEATURE EXTRACTION(a) Normalizing an ECG signal.By producing the amplitu<strong>de</strong> histogram of a completeacquisition of ECG signals, the R wave is clearly visible as being the top area of thehistogram with consi<strong>de</strong>rable mass <strong>de</strong>nsity. In figure 4.13 we show an example of an ECGsignal we acquired, where some noise is visible (left). In the middle we see the histogram ofthe signal, where the influence of the noise is limited. On the right the inverted cumulativesum is <strong>de</strong>picted; the point <strong>de</strong>tected as the R wave mean amplitu<strong>de</strong> is marked with a circle.To compute the R wave reference amplitu<strong>de</strong> for each user, we use the histogram of thesignal to create an algorithm with some immunity to noise. We <strong>de</strong>fine the vector c as theamplitu<strong>de</strong> histogram of the input ECG signal, e, such that c(a) is the histogram value foramplitu<strong>de</strong> a.∑ t=nst=t iWe use an inverted cumulative sum of the ECG signal histogram, <strong>de</strong>fined as icumsum(t i )=c(t), and search for the first point in the vector higher than a threshold (β). Thecorresponding amplitu<strong>de</strong> at this point is <strong>de</strong>fined as R ref , the R wave reference amplitu<strong>de</strong>.This threshold β was easily tuned for all users, so that the <strong>de</strong>tection algorithm correctlyi<strong>de</strong>ntifies most of the R waves of the collected data.Figure 4.13: Histogram of a noisy ECG signal. On the left the ECG signal. In the middlethe filled black area presents the histogram. A dashed line represents the histogram scaled10 times. On the right the solid line represents the inverted cumulative sum and a scaledversion to i<strong>de</strong>ntify the point where the R reference wave is <strong>de</strong>tected (marked with a circle).We assume that noise exists in only a fraction of the entire data collection so that when

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!