In case of decomposition of EEG signal into subbands to extract features, in general case which one gives more significant information ? wavelet decomposition or wavelet packet ? and why?
Dear Kazi,
No. It may or may not outperform. EEG and similar biosignals (MEG/sEGM/iEMG) spike at low frequencies. Wavelet packet decomposition (WPD) gives additional info in higher frequencies. Therefore, normal wavelet decomposition (NWD) may just work fine, if designed properly.
If I recall correctly:
// let signal=EEG from one channel
// let say decomposition level = 4
wpd (signal)=> 2^4=16 (sub-signals) 8 from low frequencies and 8 from highs
npd (signal)=> 2+1+(4-2)*1=5 (sub-signals) 2 from last layer, 1 from higher frequency first layer, and one for each middle layer
So, normal decomposition is less computationally expensive, as well.
Best regards,
J.
J. Rafiee Thanks for your explanation. Just one thing to clarify, in case of nwd, shouldn't it produce 5 sub-signals(for level 4) where 4 from lower frequency and 1 from higher frequency( in total 4 details and 1 approximation ) ?
https://www.mathworks.com/help/wavelet/ref/wavedec.html
you may want to extract information of signal in special range of frequencies. for this purpose, you may need to use wavelet packet instead of wavelet. it gives you more special information in comparison to ordinary wavelet.
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts