Wrist Angle

[dymnz]

Reports

[dymnz]

Currently working on exp 7-3. PCA version of exp 7. Compare the result of ICA and PCA and Do nothing.

old_ICA vs PCA under cross_valid new_ICA debugging

Does whitening for ICA/PCA makes sense?

FastICA whitens the input signal, can ICA be applied on RMSed signal?

• ~w/o whitening~ FastICA whitens the signal, training data amp. norm., C-V patience:
  • ICA marginally better than PCA
  • H: 32>=24>=16 @ Patience > 40/100

Data: https://github.com/dymnz/sEMG/blob/master/Reports/wang/semg_angle_est/RMS_order.md Organized: exp_6_ICA_compare.xlsx - exp_7 RMS order

[dymnz]

PRO_SUP only training effectiveness

Do this for 4-ch sEMG

See exp-10:

• Training sample can be reduced for "Full motion", though with slightly worse performance (~20%) see
  • Complex MIX, Reduced Simple/Mix Training
  • Complex MIX, Reduced Simple Training

[dymnz]

4-ch sEMG: https://github.com/dymnz/sEMG/issues/12#issuecomment-385875860

Change Matlab code sampling rate to 460 SPS

• Working on Pot.
  • Need to find the angle -> voltage mapping
  • DOF: ? degree
  • Verifying linearity for 30~150 degree, w/ 90d as metal contact pointing down and handle parallel to the ground, angle decrease as handle approaches metal contact
    • 10d increment, see change in voltage
    • Voltage from 150d to 0d inc. 10d
    • 150d: 1170
    • 140d: 1410
    • 130d: 1690
    • 120d: 1960
    • 110d: 2250
    • 100d: 2530
    • 90d: 2805
    • 80d: 3070
    • 70d: 3350
    • 60d: 3610
    • 50d: 3865
    • 40d: 3950
    • 30d: 3962
    • Linear for 150 ~ 50d, 1170 ~ 3865, checking hand DOF see if angle stay in safe range
  • 1M ohm is too large, large noise. NOW USING 50k ohm
  • Zeroing is not perfect. Repeating the same movement but the voltage won't be the same afterwards
• TODO:
  • [x] Calculate equation for voltage-angle mapping, with different 0d point
  • [ ] Calculate zeroing error
  • [x] 0d positioning in Processing
  • Set the current measurement value as 0d on hand. Simply subtract the difference between current angle and 0d.

[dymnz]

Separate NN for PRO/SUP & FLX/EXT

[dymnz]

TDSEP parameter test

For exp-10, see 'exp_10_prePro_compare'

[dymnz]

ICA, Impulse extraction,

Wrist Movements

• DOF - 1 (Pitch):
  • Flexion: Palm facing down, moving downward
  • Extension: Palm facing down, moving upward
• DOF - 2 (Roll):
  • Pronation: Forearm turning counter-clockwise
  • "pronation involves placing the palms into the prone (face-down) position, like someone would do when looking at the backs of their hands"
  • Supination: Forearm turning clockwise
  • "The motion of supination turns the palms anteriorly or superiorly to the supine (face-up) position."
  • In the anatomical position, with the arms extended to the sides of the trunk and palms facing forward, the arms are already in the supinated position.
• DOF - 3 (Yaw):
  • Radial Deviation: Palm moving counter-clockwise on the horizontal plane
  • Ulnar Deviation: Palm moving clockwise on the horizontal plane

---

Paper

• A new strategy for multifunction myoelectric control(1993)

  • ANN chosen for fast training time

• Surface EMG pattern recognition for real-time control of a wrist exoskeleton (2010)

• Simultaneous and Proportional Force Estimation for Multifunction Myoelectric Prostheses Using Mirrored Bilateral Training (2010)

  • Wrist (2-DOF) FLX/EXT & RU-DEV
  • 7-ch sEMG
  • Multilayer Perceptron
  • Feature set performance: MSV(?) < TD = TD+6AR = TD+5 wavelet marginal(TDWV)(?)
  • 7-ch semg -> TD feature -> PCA -> MLP

• Comparison of regression models for estimation of isometric wrist joint torques using surface electromyography (2011)

• SVM for Estimation of Wrist Angle from Sonomyography And SEMG Signals (2011)

  • Flexion/Extension

• Wrist angle estimation based on musculoskeletal systems with EMG (2011)

  • 4-ch, physical model
  • extensor carpi radialis longus
  • extensor carpi ulnaris
  • flexor carpi radialis
  • flexor carpi ulnaris

• EMG-based simultaneous and proportional estimation of wrist/hand kinematics in uni-lateral trans-radial amputees (2012)

  • Wrist (3-DOF), 7-ch sEMG, TDAR feature set, MLP
  • Intact-limbed:
  • (3-DOF)72.8 +- 8.29% accuracy
  • ~DoF-12 90.6% ± 3.68% (CG group, motion 1/2/4/5, accounting PROSUP/FLXEXT)~ DOF2 is R/U deviation

• Multichannel surface EMG based estimation of bilateral hand kinematics during movements at multiple degrees of freedom (2010)

  • Previous work of above
  • Intact bodied, 5 grid w/ 16-ch each (80-ch)
  • Fullwave rectified sEMG
  • 3-DOF, w/ 6 separate MLP. MLP output filted w/ LPF 1Hz(?)
  • Coefficient of determination (variability explained by estimated values)
  • FLX/EXT: 82.7 +- 2.9%
  • PRO/SUP: 76.6 +- 11.8%

• Simultaneous and proportional control of 2D wrist movements with myoelectric signals (2012)

  • 192-ch sEMG array, nonlinear transformations, Flexion/Extension/RU-Deviation

• Bayesian Filtering of Surface EMG for Accurate Simultaneous and Proportional Prosthetic Control (2014)

  • 16-ch sEMG, Flexion/Extension or Pronation/Supination and simultaneously.
  • Feature extraction (amplitude summarizing) for NMF regressor, compare MAV and BAY summarizing performance

• Quantifying Forearm Muscle Activity during Wrist and Finger Movements by Means of Multi-Channel Electromyography (2014)

  • 112-ch sEMG array

• EMG-based learning approach for estimating wrist motion (2015)

  • Based on EMG-based simultaneous and proportional estimation of wrist/hand kinematics in uni-lateral trans-radial amputees (2012)
  • 11-ch sEMG, SVR
  • Wrist angle @ different arm position
  • Circular motion of FLX/EXT & RU-Deviation
  • Same day Same pose: 78% | Same day Across pose: 63% | Different day: 50 ~ 0%

• Continuous Estimation of Wrist Angles for Proportional Control Based on Surface Electromyography (2016)

  • 4-ch sEMG, BPNN w/ fixed windowed RMS pre-processing has the best result
  • Separate tests for Flexion/Extension/Pronation/Supination
  • RMSE | WF: 10.42 | WE: 8.23 | WS: 16.53 | WP: 13.09
  • Accuracy decrease in the presence of grip force

• Continuous estimation of hand's joint angles from sEMG using wavelet-based features and SVR (2016)

  • SVR, DWT

• EMG-Based Continuous and Simultaneous Estimation of Arm Kinematics in Able-Bodied Individuals and Stroke Survivors (2017)

  • NARX model (Recurrent NN), 3-DOF shoulder/Elbow/Wrist(FLX/EXT), 6-ch sEMG
  • LPF @ 4Hz, 99%+ accuracy

• Continuous Motion Decoding from EMG Using Independent Component Analysis and Adaptive Model Training(2014)

  • Mentioned Gaussian characteristic but used ICA???

May Be Relevant:

• Neural network committees for finger joint angle estimation from surface EMG signals
• Real-time simultaneous and proportional myoelectric control using intramuscular EMG (2014)
• Estimation of finger joint angles from sEMG using a recurrent neural network with time-delayed input vectors (2009)

• *Extracting Simultaneous and Proportional Neural Control Information for Multiple-DOF Prostheses From the Surface Electromyographic Signal

Wrist movement muscle:

• Extension | Should find which muscle other experiments are focusing on

  • Source | Highlighted because used in other paper
  • Extensor Digitorum <- testing
  • Extensor Carpi Radialis Longus
  • Extensor Carpi Ulnaris <-- Chosen
  • Extensor Pollicis Longus
  • Characteristic
    • POWER: Linear transform of EMG
    • RMS: sqrt(POWER), non-linear
    • LOG-VAR: log(POWER), non-linear
  • Use non-linear transform to linearize relationship between Feature and Joint angle
  • Linear regression performance better w/ RMS, LOG-VAR
  • MLP is better w/ POWER
  • Linear regreesion w/ LOG-VAR has best performance

• Flexion | Should find which muscle other experiments are focusing on

  • Source | Highlighted because used in other paper
  • Flexor Carpi Radialis <-- Chosen
  • Flexor Carpi Ulnaris
  • Flexor Digitorum Superficialis <- testing
  • Flexor Pollicis Longus (Thumb movement)

• Pronation | Source | Highlighted = chosen

  • Pronator Teres: Near elbow
  • Pronator Quadratus: Near wrist (Hard to locate)

• Supination | Source | Highlighted = chosen

  • Supinator Muscle : Near elbow (Hard to locate)
  • Biceps Brachii (Tried and failed)

[yenhuai77]

thx