When using the ECG signals for detecting heart anomaly, is it necessary to have high resolution sampled signals or lower resolution sampling can be accurate enough?
This is a very good question. The general answer does not exist. The minimal sampling frequency strongly depends in the ECG feature that you are measuring and on the methods uses to detect it.
What I usually do is to use the maximal possible sampling frequency to study the subject. When the conclusions are made, it is possible to gradually decrease the sampling frequency by omitting some from the original data.
You observe the change of your results and estimate the minimal sampling frequency that does not spoil the results.
It is simple and you are sure that it gives you the right estimate. To make your estimate even more robust, you can acquire data from a different source. Additionally, you can try to add some noise to the original data with a predefined distribution. It this way you can test the robustness of your method against various influences.
I would follow the recommendation from AHA, mentioned by Mr. Pala; One reason ring a bell in my mind rely on the fact that it's hard to design a simple digital filtering with a low cutoff frequency; once the sample frequency of the signal itself is too high.
I agree with Prof. Jiří Kroc it depends in the ECG feature that you are measuring. In some cases using lower resolution sampling lead to omit some parts of the original signal. So, if you need to work on all parts of the signal I think you should use high resolution sampled signals. But if you working on specific features and you sure that these features not in the omitted part, I think you should use lower resolution sampling.
There is another big issue with the detection of heart anomalies, features, frequency, and other issues. Safety. Safety of detection. Sensitivity and specificity of detected outputs of techniques, analytical methods, etc. must be kept as high as possible.
We need some internationally recognized, standardized database of ECG recordings of all possible scenarios that enable testing of our software against the standard. Sensitivity and specificity of all measured features can be easily compared among all software types and proves.
It is very important for safety I people and patients wearing such devices and using this software.
To produce any waveform you must sample it with a frequency fs greater equal the twice the highest frequency fm contained in it. fs>= fm.
Concerning ECG if you to reproduce its pulses you have to analyze the pulse in the frequency domain by following the practical approach followed by Jiri.
It is so that you choose a high sampling rate and then do the analysis to get the FFT spectrum and then reproduce your signal by IFT
Reduce fs slowly and repeat the same till you could not reconstruct the wave form this is the minimum sampling frequency.
Another very practical method is to enlarge the pulse and then estimate the number of regular points that can be used to draw it shape. The time between the adjacent samples will be the sampling time.
it depends on how much features that you want to process. the extreme minimum sampling freq. is at least 100 Hz. The maximum freq. of ECG is about 150 Hz, so for appropriate sampling freq. should be Fs>=Fm.