sEMG Preprocessing/Feature about semg HOT 8 OPEN

Wavelet

• Surface Electromyography Signal Processing and
  Classification Techniques
  • Wavelet denoising

from semg.

PCA Signal Filter and Reconstruct

Basic idea: Raw signal -> PCA -> discard unwanted -> reconstruct -> clean signal

• PCA maximize variability
• 4-ch sEMG for 4 muscle, each channel should have more or less equal variability
• Discarding a PCAed signal would remove some variability from all signal, equally
  --> may not be suitable

Look for more example

from semg.

ICA

• sEMG signal has Gaussian dist., not suitable for ICA
• Is RMS signal suitable for FastICA? probably not
  Dropping ICA, at least temporarily. See 'Meeting 2018/05/11 - 4-ch SEMG'

• A FastICA Algorithm for Non-negative Independent Component Analysis
• Algorithms for Nonnegative Independent Component Analysis
  • Find the latest iteration of this

from semg.

Alternative BSS

See #42 for TDSEP

• Exploiting time separation. ICA/PCA lack temporal information
  • signal separation nonstationary
  • signal separation time
  • signal separation temporal
  • temporal decorrelation
  • temporal structure BSS

• A time–frequency blind signal separation method applicable to underdetermined mixtures of dependent sources

• Nonstationary source separation using sequential and variational Bayesian learning

• Adaptive filtering for music/voice separation exploiting
  the repeating musical structure

• Temporal FastICA

from semg.

Pulse Gen Integration

• Pulse mode seems be performing better w/ separated FLX/EXT or PRO/SUP

from semg.

Why BSS

X = AS, Measured EMG
Y = f(S), Angle

S: EMG
A: Mixing matrix
f: Non-linear function
---
NN: 
Have Y and X
Find inv(A), to get S = inv(A)X
then, find f(), to get Y = f(S)

NN w/ BSS:
Have Y and X
BSS: Find inv(A)
NN: Only needs to find f(), simplify NN's job

from semg.

New

• RMS -> ICA
  • Done previously, check history
• RMS -> TDSEP
  • Makes more sense then above

from semg.

Nonnegative Matrix Factorization (NMF)

• Nonnegative Matrix Factorization for the Identification of EMG Finger Movements: Evaluation Using Matrix Analysis
• Decomposing Time Series Data by a Non-negative Matrix Factorization Algorithm with Temporally Constrained Coefficients

General info on PCA/ICA/NMF:
https://pdfs.semanticscholar.org/presentation/3389/7f174795bb193a4b91d65b767ade629f1f6d.pdf

from semg.