I am working on a project related to sEMG classification. In order to process the signal, I am trying to filter the signal, especially power interference. The signal is acquired from Delsys Bagnoli 16, and the sample rate was 4000. I collected 6 signals simultaneously, and some signals showed this abnormal behaviour in fft (see the figure, zoomed to 0-700Hz). Therefore, I tried notch filters to remove the spikes, but it seems like it has other spikes around 50Hz. For example, see the next figure. I need to know the reason for this behaviour. Is it due to fft calculation? (see the last figure, Matlab code. it was recommended by the Matlab (https://www.youtube.com/watch?v=VFt3UVw7VrE , at 5:17)) or due to a filtering problem at the amplifier?
A possibility: in many countries, the power supplied from an outlet is alternating current at 50Hz or 60Hz. Therefore, many electrical devices in your lab. (eg, fluorescent lamps) generate noises at that frequency and your circuit receives them.
Check this link, I was facing this problem when I was designing amplifier.
Different factors are discussed in this question.
https://www.researchgate.net/post/How_to_remove_50Hz_noise_from_DC_signal
Kengo Shibuya Thank you for replying. I am aware that this is due to the AC current but there are other spikes in the fft like 49.1Hz, 50.3Hz. I am looking for the reason for that.
There are two reasons for the behaviour that you are seeing. First is the erratic phase noise of main electricity. You know, mains electricity is not stable 50 Hz. It is affected by varying load, but even the source of the power is not a steady stream of energy: traditionally, and still in the most common case it is based on mechanical rotation that is transformed to electricity. So, within a short period of time you are not seeing a smoothly descending phase noise, but erratic spikes near 50 Hz, and the main peak is hardly ever exactly 50 Hz.
Second reason for some of the "spikes" that you are seeing is that you are not actually seeing the exact spike but sinc-response of the FFT because the length of the measurement is not an exact multiple of the sine wave. You can see the effect of this by comparing figures:
> plot([0:4e5-1]/4e5*4000,20*log10(abs(fft(sin(2*pi*50*[0:4e5-1]/4000)))))
> plot([0:4e5-1]/4e5*4000,20*log10(abs(fft(sin(2*pi*50.005*[0:4e5-1]/4000)))))
50 Hz spike: Ac electric frequency;
Other spikes (in multiples of 50 Hz): Ac electric frequency harmonic;
All generated by Ac power supply.
And for 49.1 Hz or 50.3 Hz, I may suggest you study this key word: side band, which I think, is generated by modulation of sampling frequency and other frequency components.
A widespread phenomenon that might not necessarily contain specific sEMG information.
In the attached work "A wavelet-based method for power-line interference removal in ECG signals" (available in https://www.scielo.br/scielo.php?pid=S2446-47402018000100073&script=sci_arttext) this problem is solved using the wavelet transform, which is more efficient than the notch filter.
Bruno Rodrigues de Oliveira Thank you for the information. I need to implement this filtering system for real-time application. Also, my window size for the calculating features would be 400 data points (100ms, 4000Hz sampling rate). Can you elaborate more on this aspect? Do you recommend the wavelet-based method for this kind of short signal pulses?
Xiaolong Wu Thank you for the information and suggestion. It helped me to understand the problem well. Thank you again.
Dulanjana Perera I recommend the wavelet transform (DWT), because is a filtering method very fast, ideal for real-time applications. The approach mentioned in https://www.scielo.br/scielo.php?pid=S2446-47402018000100073&script=sci_arttext is even faster, as it does not implement threshold functions.
- Badges
- Science topic
- Similar topics
- Control Systems Engineering
- Filtering