BIO – BSA3
Exercises for Part 2 (Signal Analysis) Exercise #06 – Feature Extraction
1.) Showing RSA by R-peak feature extraction:
In exercise 5.1 the existence of the RSA could be demonstrated by sFFT. A better method for obtaining the heart rate is inverse R-R distance.
a. Use any ECG of your choice (BIOPAC-recording, ECD-database, internet) and develop a simple algorithm which detects the R-peaks of the ECG (if necessary: do all required pre- processing tasks)
b. Visualize the location of each R-peak
c. Calculate the duration auf each RR-interval (= period of each heart beat)
d. The inverse RR-period is the frequency (heart rate) which corresponds to each heartbeat.
Calculate the HR for each heart beat.
e. Plot the heart rate over time.
f. Find the RSA-frequency from the plot.
2.) Measurement of periodicity: Compare different methods to extract the periodicity in an ECG signal (Peak to peak analysis, FFT Analysis, Autocorrelation)
a. Compare the three measures for a 5 minute sample of ECG recordings (use any ECG you like)
b. Add a different amount of artificial noise and try to figure out which method is most robust with respect to noise
3.) QRS Annotation:
a. Use any ECG of your choice (BIOPAC-recording, ECD-database, internet). Perform all
required pre-processing so that Q, R and S-peak are clearly visible in the ECG. Use at least a
recording length of 2 min.
b. Split the ECG in to half. By only using the first half, develop an algorithm to detect the Q, R
and S-peak. Use as few “magic numbers” as possible.
c. Test the algorithm individually on the first half (training data) and the second half (test
data). Visualize your results.
Task
1
2
3
Sum
Points
3
2
3
8