emg acquisition and myoelectric control strategiesusers.wpi.edu/.../l20-emglecturewpi.pdf ·...

EMG ACQUISITION AND MYOELECTRIC CONTROL STRATEGIES

MICHAEL TWARDOWSKI,

RBE PHD STUDENT

De Luca CJ, The Use of Surface Electromyography in Biomechanics,

J. Applied Biomechanics, 13: 135-163, 1997

“Electromyography is too easy to use

and

consequently too easy to abuse”

What is EMG?

[s]370 372 374 376 378 380

[Volts]

-0.002

-0.001

0

0.001

0.002

muscle fibers

Electro – Electric

Myo – Muscle

Graphy – Graph

“Electromyography (EMG) is the

study of muscle function through

the enquiry of the electrical signal

the muscles emanate”

Muscles Alive

Basmajian & De Luca, 1985

4copyright © 2007 Delsys Inc. ISBN: 978-0-9798644-0-7

The surface EMG (sEMG) signal can be used in the following applications

Obtain parameters from individual or groups of muscles

• Force Activation Time (ON – OFF)

• Fatigue

Compare behavior of different muscles

• Relative amount of contribution

• Co-activation

• Pattern identification (tasks)

Why Use the sEMG Signal?

Revealing and harnessing

the neural control of

movement

How the Brain Controls Muscles…

Neural Control

Nervous System

The spinal cord is the major pathway of the nervous system.

Motor Unit

Motorneuron: neuron in

the spinal cord that

connects to muscles.

Motor Unit:

Motorneuron, its axon

and all of the muscle

fibers that in connects to.

It is the basic functional

unit of the

neuromuscular system.

Motorneuron Communication

Neuron Action Potential:

• change in electrical

potential

• produced by flow of

molecules in/out of cell

• propagates in one

direction along cell

• Means of neural

communication

Motor Unit Mechanical Output

De Luca CJ & Erim Z. Common drive of motor units in regulation of muscle force. Trends in Neuroscience, 17: 299-305,

1994.

Twitch Tetanic ForceMU firings

Spinal Cord

Motor neurons

Synapse

Muscle fibers

copyright © 2007 Delsys Inc. ISBN: 978-0-9798644-0-7

Drive

to MUs

De Luca CJ and Erim Z. Common drive of motor units in regulation of muscle force. Trends in Neuroscience, 17: 299-305, 1994.

Noise 1

Motor Unit Action

Potential

MU 1

EMG Signal

MU 2

MU N

Noise 2

Noise NFiring MUAPT

+

+

+

+

Motor Unit Control and EMG Signal

sEMG Signal Acquisition

EMG Signal

• System subtracts two signals:

(m1+n) – (m2+n) noise is removed

• Resultant EMG signal:

(m1-m2) is amplified

Recording

copyright © 2007 Delsys Inc. ISBN: 978-0-9798644-0-7

• The quality of the sEMG signal should be the first concern of any tests performed to collect sEMG signals.

• The quality of the EMG signal depends on:• Sensor characteristics

• Sensor Location

• Electrode-skin interface

• Noise contamination

• Cross-talk from other muscles

Signal Quality

copyright © 2007 Delsys Inc. ISBN: 978-0-9798644-0-7

Sensor Characteristics

• Small inter-electrode spacing• sensors for small muscles

• selective recording from individual muscles

• reducing cross-talk

• Differential amplification• removing ambient noise (50, 60 Hz, Radio frequencies)

• Fixed spacing of electrodes for consistent signal properties• Amplitude and frequency of signal is a function of the distance

between the electrodes

contoured skin-

electrode interfacedEMG System

for MU studies

Trigno™

EMG + ACC

Ag/AgCl

Electrodes

High Density EMG

copyright © 2007 Delsys Inc. ISBN: 978-0-9798644-0-7

StancePhase

SwingPhase

SwingPhase

SwingPhase

StancePhase

Sensor Characteristics

10mm DD bar

• Crosstalk reduced at 10 mm spacing• Crosstalk substantially reduced for 10 mm Double Differential

Double

Differential

10mm

Crosstalk

muscle

22mm -

+

-

+

-+

-+ -

+

Single

Differential

Single

Differential

10mm

De Luca CJ. The use of surface electromyography in biomechanics. Journal of Applied Biomechanics, 13: 135-163, 1997.

Innervation

Zone Proper

sensor location

• Increases the signal

• Increases the signal

to noise ratio

• Reduces cross-talk

Sensor Location:

Easiest, cheapest, most effective way of obtaining a better

signal to noise ratio

copyright © 2007 Delsys Inc. ISBN: 978-0-9798644-0-7

• Contoured surface maintains proper electrical contact with the skin during movement

Electrode Skin Interface

copyright © 2007 Delsys Inc. ISBN: 978-0-9798644-0-7

Baseline noise + EMG signal

Noise Contamination of the EMG Signal

50 100 150 200 250 300 350 400 450 500

Frequency

0.2

0.8

0.6

1.0

0.4Po

wer

20

0

Baseline noise = 1.2 uV RMS sEMG signal (Delsys sensors)

copyright © 2007 Delsys Inc. ISBN: 978-0-9798644-0-7

• Physiological Noise• EKG, EOG, respiratory signals, etc.

• Ambient Noise• Power line radiation (50, 60 Hz)

• Cable motion artifact

• Baseline Noise• Electro-chemical noise (skin-electrode interface)• Thermal noise (property of semi-conductors)

• Movement Artifact noise• Movement of electrode with respect to the skin (induced

by force transients or movement of the skin)• This is the most obstreperous noise

Noise Contamination

copyright © 2007 Delsys Inc. ISBN: 978-0-9798644-0-7

• Reduce Baseline noise by preparing the skin • Clean skin with alcohol

• Remove hair

• Peel off top dead layer of skin with hypoallergenic tape

• Use state-of-the-art active sensors

Baseline Noise

copyright © 2007 Delsys Inc. ISBN: 978-0-9798644-0-7

• Induced by force transmission through the muscle and skin

• Caused by relative movement of sensor with respect to skin• Poor electrical contact between electrode and skin• The electrolyte material between electrode and skin

• More dominant in low level EMG signals

• Contaminates EMG signal• Frequency components superimpose on those of EMG signal

• How to Reduce Movement Artifacts• Skin Preparation• Not using electrolyte gel• Filtering

Movement Artifact

copyright © 2007 Delsys Inc. ISBN: 978-0-

9798644-0-7

Baseline noise + EMG signal + Artifact

50 100 150 200 250 300 350 400 450 500

Frequency

0.2

0.8

0.6

1.0

0.4

0

Po

wer

-2.5

-1.5

-0.5

0.5

1.5

2.5

0.2 0.4 0.6

Time (s)

Accele

rati

on

(g

)

Movement Artifact

Noise Contamination of EMG Signal

20

copyright © 2007 Delsys Inc. ISBN: 978-0-9798644-0-7

EMG signal with Baseline noise and Artifact reduced

0.2

0.8

0.6

1.0

0.4

0

Po

wer

FILTER 20- 450 Hz

Filtering the sEMG Signal

50 100 150 200 250 300 350 400 450 50020

Frequency

On/Off and Proportional Control of Prostheses

AMPLITUDE ESTIMATION

Sliding RMS (20 ms) Sliding RMS (250 ms)

Commercially available control schemes require unintuitive, unnatural sequences of

contractions to control multiple DOF.

Pattern Recognition Control

PATTERN RECOGNITION

Li, G., & Kuiken, T. A. (2009). EMG pattern recognition control of multifunctional prostheses by transradial amputees.

FEATURE EXTRACTION

Phinyomark, A., Phukpattaranont, P., & Limsakul, C. (2012). Feature reduction and selection for EMG signal classification. Expert Systems with Applications.

• Amplitude Estimates (Signal Energy)

• Root Mean Square

• Mean Amplitude Value

• Signal Complexity

• Waveform Length

• Slope Sign Changes

• Zero Crossings

• Frequency

• Willison Amplitude

• Prediction Model

• Autoregressive coefficients

• Time Dependencies

• Mean Absolute Value Slope

NON LINEARITY sEMG VS. FORCE

Investigations have shown that the amplitude of the sEMG

signal is nonlinearly related to the actual force of the

contracting muscle.

Lawrence, J. H., & De Luca, C. J. (1983). Myoelectric signal versus force relationship in different human muscles. Journal of Applied Physiology .

Force

RM

S (

MV

C %

)

0 20 40 60 80 0 20 40 60 800 20 40 60 80

Implanted Electrodes

Traditional Intramuscular Recording Techniques

INVASIVE EMG

Motor unit action potentials (MUAPs) can

be recorded with the use of needle EMG.

The tip of the needle is the recording

electrode.

Another method is fine wire EMG. The needle

has a thin wire in it with a hook at the end.

Once inserted, the needle is removed and

only the wire remains in the muscle.

IMES

Implantable Myoelectric Sensors

• Control of 1 DOF for every 2 sensors

IMPLANTABLE DRAWBACKS

Implantation and maintenance of invasive neural interface:

• disturb the integrity of the residual nerves or muscles

• pose great health risks

• present substantially greater health costs

Intense fear and mutilation anxiety associated with surgery

THE INCEPTION

OF A LEGACYEMG Decomposition

First conceived and implemented by Carlo J. De Luca in 1982.

By 2000s developed non-invasive surface dEMG.

[Insert 20s video clip of decomposing EMG signal]

36

. . . . . . . . .

50 uV

0

5

10

15

20

25

30

35

40

Forc

e (%

MV

C)

12

34

5

6

7

8

9

39

40

41

42

43

4445

46

47

Time (s)

Mo

tor

Un

it N

um

ber

6 84 504846

Decomposition

Raw sEMG Signal

dEMG

Force Gradation

Recruitment

Threshold,

RT

Firing Rates

Force

Motor Unit Firing TrainMotor Unit Action Potential

0 5 10 15 20 250

10

20

30

40

50

Force (%

MV

C)

Time (s)

5

10

15

20

25M

ean

Fir

ing

Rat

e (p

ps)

50Recruitment Threshold (%MVC)

10 20 30 40

10

5

15

20

25

0

0

y = -0.15x + 22

De Luca CJ, LeFever RS, McCue MP, and Xenakis AP. Behavior of human motor units in different muscles during linearly-varying contractions. Journal of Physiology, 329: 113-128, 1982.De Luca CJ and Erim Z. Common drive of motor units in regulation of muscle force. Trends in Neuroscience, 17: 299-305, 1994

Onion Skin

Fo

rce (

%M

VC

)

Common Drive

25

20

15

10

5

00 5 3510 3015 20 25 40

0

13

26

52

39

Time (s)

Mea

n F

irin

g R

ate

(p

ps)

0-0.1 0.1-0.5 0.5-1 1

-1

1

0

Time

Lag (s)

Cro

ss C

orr

ela

tio

n

30

)()()(

tyxtc

Cross-Correlation Function

UNDERSTANDING MOTOR CONTROL

Motor unit control properties

Correlated control of motor units

Biomechanical benefits of control schemes

dEMGdEMG

Conceptualize new

EMG technology

Development of

EMG product

Advance understanding

of human movement.

Enabling new

research applications

From the NeuroMuscular Research Center

Scientific/Research

Questions

Trigno™ Mini

SensorThe Muscle

Fatigue Monitor

1st commercial

Signal Quality

Monitor

Trigno™ EMG

Wireless Sensor

A Foundation of Innovation in EMG Technology

Supporting today’s innovations

Trigno™ IM and dEMG

1979 1982 1985 1997 2000 2006 2008 2013

Marketed the stand-

alone EMG sensor

Double-Differential,

cross-talk reduction

contoured skin-

electrode interface

dEMG System

for MU studies

Trigno™

EMG + ACC

Trigno™ EMG

+ IMU SensorFirst fixed-spacing

surface sensor

TODAYS UNIVERSITY TALENT SUPPORTS

TOMORROWS INNOVATIONS

Training Researchers for Over 40 years

25 MS Theses, 12 PhD Dissertations, 41 Research Associates

Internship Programs FUTURE EMPLOYEES

Research Fellowship Programs FUTURE EMPLOYEES

CONCEPTS THAT WILL SHAPE OUR FUTURE

Investigating

EMG for

Subvocal

Speech

• AAC – Augmentative and

Alternative Communication

• Does not rely on acoustic

excitation

• Utilizes subvocal (mouthed)

speech

EMG-BASED SUBVOCAL COMMUNICATION

The alternative to voiced speech

Signal

Processing

EMG from multiple muscles

processed concurrently Parameterized data delivered

to back end algorithms Recognition results and error

rates for each utterance

1…2…

3…

4…

nAI

Recognition

Algorithms

Raw

Data

Hypothesis

Testing

SUBVOCAL ALGORITHM ARCHITECTURE

Intelligent Signal Processing + Strategic Machine Learning

- TIMIT standardized

speech data corpus

- Frequency-phoneme

combinations to test

unforeseen words

- State detection of

Start/End of speech

- Separation of speech and

non-speech muscle activity

- EMG features of speech

- LDA Feature reduction

- HMMs based on acoustic

speech processing

- Deep Learning

(experimental)

- Word error rates

- Subject specific

parameter

performance

APPLICATIONS OF SUBVOCAL TECHNOLOGY

Consider the alternatives to voiced speech

Subvocal

prosthesis

for speech

impaired

Subvocal

systems for

stealth

speech

Subvocal

speech in

quiet

environments

Subvocal

speech

translation

Subvocal

speech for

noisy

environments

Tracking

Movement

Disorders

CONCEPTS THAT WILL SHAPE OUR FUTURE

TRACKING THE HEALTH OF HUMAN MOVEMENT

• Consider Parkinson’s Disease,

Essential Tremor, etc.

• Movement disorder changes

throughout the day

• Tracking symptoms is important

for treatment

• But current monitoring methods

are unreliable

SENSOR SYSTEM FOR MONITORING MOVEMENT

Automated real-time detection

TrignoTM EMG + IMU Sensors

- 1 Ch EMG,

- 9 Ch IMU (Acc, Gyro, Mag)

- synchronous sensors

- wireless transmission

SENSOR SYSTEM FOR MONITORING MOVEMENT

Proposed System

Movement Disorder Results

Data Acquisition System Platform-Generated Applications

Data Logger & Sensor

Essential Tremor

Parkinson’s

NormalFreezing

7%

Dyskinesia20%

Tremor33%

Summary Status

Other….

Tablet