Docstoc

Microsoft PowerPoint - SEMG for

Document Sample
Microsoft PowerPoint - SEMG for Powered By Docstoc
					    Hand gesture for HCI using
           ICA of EMG

Ganesh. R. Naik, Dinesh. K. Kumar, Vijay Pal Singh.
    School of Electrical and Computer Engineering
    Royal Melbourne Institute of Technology
    University Melbourne

M. Palaniswami

   The Department of Electrical and Electronic Engineering.
   The university of Melbourne, Melbourne


                      ganesh.naik@rmit.edu.au                 1
       Presentation Outline
Introduction

Electromyogram

Independent Component Analysis

ICA for Identifying hand gesture

  Methodology
  Results and observations
  Discussions

Conclusion and future work
                  ganesh.naik@rmit.edu.au   2
 Hand Gesture Identification.

Hand Gesture Identification has various Human
Computer Interaction applications. E.g. controlling
machines and computers.

Primarily used systems

  Vision based system

  Mechanical sensors

  Electromyogram


                    ganesh.naik@rmit.edu.au       3
            Electromyogram
Electromyography (EMG) is a result of the spatial
and temporal integration of the motor unit action
potential (MUAP)
It is typically the only in vivo functional examination
of muscle activity used in the clinical environment.
Recorded Non invasively
Used for Dynamic measurement
Applications:
  Sports, training, rehabilitation, machine and   computer
  control



                    ganesh.naik@rmit.edu.au              4
    Independent Component
           Analysis
Blind Signal Separation (BSS) or Independent
Component Analysis (ICA) is the identification &
separation of mixtures of sources with little or no
prior information.


Applications include:

  Audio Processing
  Medical data
  Finance
  Array processing (beam forming)
  Coding
                   ganesh.naik@rmit.edu.au        5
                               ICA or BSS
Independent                               Observed                                   Recovered
components                                sequences                                 independent
                                                                                    components




                                                               De-mixing
                                                                process




                                                …
  …




                                                                   W

                     Mixing
                     process
                                A
Blind
Source



Source// Te Won Lee, “Independent component analysis, theory and applications” Kluwer Academic
        Publishers, MA, USA, 1998.
                                    ganesh.naik@rmit.edu.au                                      6
ICA for Identifying Hand Gesture
  Our method

  hand gesture identification using prior knowledge
  of the muscle anatomy

  Model based approach that provides a well defined
  muscle activity pattern




                  ganesh.naik@rmit.edu.au         7
ICA for Identifying hand gesture
Methodology
Three hand actions were performed and repeated 15 times.
There was no external load.
Electrodes over the skin of the forearm were placed as shown
in the table below:

 Channel               Muscle                          Function
    1      Brachioradialis                        Flexion of forearm
                                                  Adduction and
    2      Flexor Carpi Ulnaris (FCU)
                                                    flexion of wrist
                                                  Abduction and
    3      Flexor Carpi radialis (FCR)
                                                    flexion of wrist
                                                  Finger flexion while
           Flexor digitorum superficialis
    4                                                avoiding wrist
              (FDS)
                                                     flexion
                        ganesh.naik@rmit.edu.au                          8
            Experimental setup




    Delsys Electrode
source// www.delsys.com)
                                     Electrode Placement during the
                                     conduction of the experiment

                           ganesh.naik@rmit.edu.au                    9
              Methodology
The actions are listed below:
  Wrist flexion (without flexing the fingers).
  Finger flexion (ring finger and the middle finger
  together without any wrist flexion).
  Finger & wrist flexion together but normal along
  centre line

The hand actions and gestures represented low level
of muscle activity The hand actions were selected
based on small variations between the muscle
activities of the different digitas muscles situated in
the forearm.

                    ganesh.naik@rmit.edu.au          10
                  Analysis.
recorded signals (x) were analysed using Fast ICA
matlab software package with the following.

                    x = As
where ‘x’ is the recorded data, A mixing matrix and
‘s’ is the sources.

The sources are recovered using the following.

                    s = Bs

where ‘B’ is the inverse of matrix ‘A’.


                   ganesh.naik@rmit.edu.au       11
           Analysis                                                  continued……
         Four Chanel EMG                                              Estimated Sources
         Recordings                                                   using ICA
                  Mixed signals
0.05                                                       10

   0                                                        0

-0.05                                                      -10
     0    500   1000         1500   2000      2500               0     500   1000   1500   2000        2500
 0.02                                                       5

   0                                                 ICA    0

-0.02                                                       -5
     0    500   1000         1500   2000      2500               0     500   1000   1500   2000        2500
  0.5                                                       5

   0                                                        0

-0.5                                                        -5
    0     500   1000         1500   2000      2500               0     500   1000   1500   2000        2500
0.06                                                        5

0.04                                                        0

0.02                                                        -5
    0     500   1000         1500   2000      2500               0     500   1000   1500   2000        2500




                                           ganesh.naik@rmit.edu.au                                12
   Analysis                           continued……

The RMS values were computed as follows:

                           N
                     1
                         ∑s
                                  2
                                  i
                     N    i =1


RMS result yields four sources Sa, Sb, Sc and Sd

RMS Values for the hand actions were trained and
tested with Back propagation neural network



                  ganesh.naik@rmit.edu.au           13
                  Results:

   Action        Action identified for each sub
 performed                experiments
Wrist flexion   100% 100% 100% 100% 100%


  Finger        100% 100% 100% 100% 100%
  Flexion
Finger flexion 100% 100% 100% 100% 100%
  and wrist
   flexion


                   ganesh.naik@rmit.edu.au        14
 Discussions and conclusion

A new approach that combines semi-blind ICA along
with neural networks has been used to separate and
identify hand gestures.

The results demonstrate that the technique can be
effectively used to identify hand gestures based on
surface EMG when the level of activity is very small.




                   ganesh.naik@rmit.edu.au         15
              Future work

Test the above technique for different muscle
activities

Test the technique for inter day-inter subject
variation.

Automate the semi-blind operation




                  ganesh.naik@rmit.edu.au   16
Questions?


  ganesh.naik@rmit.edu.au   17
Thank You



  ganesh.naik@rmit.edu.au   18

				
DOCUMENT INFO
Shared By:
Categories:
Stats:
views:28
posted:9/21/2010
language:English
pages:18