Dear all,
I am using the pulse sensor, a low cost device for PPG detection in the index finger
https://pulsesensor.com/
Although I am quite still, the detected signals are heavily corrupted with low-frequency and high-frequency noise . Which simple and robust signal processing technique do you recommend for
solving this case? If it works in real time, much better.
I need the clean PPG on order to measure accurately the heart rate (HR) and the heart rate variability (HRV).
Thank you all!
Fernando
Usually, the most straightforward and robust way is use of Fourier transform.
In the simplest variant you acquire data and do interleaved FFT, in the spectra it is usually quite easy to separate signal and interference.
You can have real-time (sliding window Fourier) which is not based on FFT approach and would demand a bit of computational power.
There are various denoising methods: methods, based on Fourier transforms, including improved ones:
Article One Technique to Enhance the Resolution of Discrete Fourier Transform
as well as wavelet-based techniques:
Article Analysis of the Quantization Noise in Discrete Wavelet Trans...
and empirical mode decomposition:
Conference Paper Denoising Algorithm Based on EMD with Adaptive Adjustment of...
Article Empirical Mode Decomposition for Signal Preprocessing and Cl...
We considered EEG signals once, they can be even more noised than PPG. Some denosing techniques were investigated in:
Article Filtering Techniques for Chaotic Signal Processing
In real-time applications, fine tuned FIR filters can perform well. Anyway, feel free to contact me if you need more help.
Fernando A. Marengo Rodriguez
If you need real-time filtering, I would suggest a Kalman filter approach to your problem.
Article Improved elimination of motion artifacts from a photoplethys...
An optimal filter for short photoplethysmogram signals (https://www.nature.com/articles/sdata201876) with Matlab source code.
Good, thank you ALL for your kind answers! I will test your suggestions! Have you tested them by yourselves?
Fernando, it depends on what you are going to use the PPG signal for. Is phase important? (deley is -d(phase)/dw, so if you have linear phase you have flat delay, which is good for morphology preservation. If yes, then IIR filters are not a good choice because, although they are 'economical' in terms of order, they tend to have very non-linear phase, especially close to the cut off frequency. An FIR can be constructed with linear phase (provided the coefficients have even symmetry around a centre point. Of course you need a larger order of an FIR than an IIR for the same sharpness of the filter, but such is life.
The most important question is, what is the cut-off frequency you will choose. Think about that carefully. (I do not want to give you all the answers, but invite you to think and point the way). :-)
Ah, and yes, I have used filters all my career (and I am 63 now, so, a few years' experience we are talking about), inclusive for PPG signals.
Best of luck!
- Badges
- Science topic
- Similar topics
- Methodology
- Research Methodology
- Research Papers