Hello everyone!
Through my studies, I used a lot of signal analysis methods for medical data (mostly RR interval series), focusing on the nonlinear ones such as:
- Shannon entropy,
- sample entropy (https://journals.physiology.org/doi/full/10.1152/ajpheart.2000.278.6.H2039?view=long&pmid=10843903),
- approximate entropy (
Article Approximate entropy (ApEn) as a complexity measure
) - detrended fluctuation analysis (Peng et al. 1994,
Article Mosaic organization of DNA nucleotides
), - multiscale multifractal analysis (
Article Multiscale multifractal analysis of heart rate variability r...
) - symbolic analysis (
Article Assessment of cardiac autonomic modulation during graded hea...
)
Currently I'm working on RR interval series obtained during listening to (or playing) short excerpts of music pieces. I'm wondering which nonlinear method would be the most appropriate for short-term data, from 30 seconds to 5 minutes (it's about 30-500 samples per signal). My preliminary results showed that I see significant differences between the baseline and the music piece period for Shannon entropy (this parameter works much better than most linear indices). In turn, I cannon see any interesting results using sample entropy and I think that these signals are too short for this method. Similarly, DFA cannot be used for a such short period.
My question is, what other nonlinear methods can I use for short-term analysis and maintaining a good quality level of the results?
I will be grateful for any suggestions.
Best
Mateusz Solinski
Hello Mateusz,
you could try recurrence analysis. It's been used for heart rate variability data for more than 20 years (see Marwan et al. (2002)) and offers a broad range of recurrence quantification measures. It's quite popular for short nonlinear signals. For implementations, have a look here: http://www.recurrence-plot.tk/programmes.php
Hope this helps.
Best,
Tobias Braun
Hello Tobias,
Thank you! Yes, I think recurrence analysis would be useful. I've been not aware that it has been used for short signals as well.
What about extension of information theory and using of symbolic dynamics?
Hi Mateusz
according to Tobias, I suggest you the Recurrence Quantification analysis,
you could start with time series with a comparable amount of points.
Ciao
Giovanna Zimatore
Hi Mateusz,
That's a very interesting discussion. I would also try "fuzzy entropy", as it is not dependent on the number of matches (doi: 10.1109/TNSRE.2007.897025), as well as DFA for a short and narrow interval of window sizes (say 5-10).
Best,
Luiz
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts