seminar topics

Upload: safu117

Post on 12-Oct-2015

62 views

Category:

Documents


1 download

DESCRIPTION

Ieee conference papers

TRANSCRIPT

  • CONFERENCE ON SIGNAL PROCESSING AND REAL TIME OPERATING SYSTEM (SPRTOS) MARCH 26-27 2011

    COPO306-1

    Real-time Hand Gesture Recognition Techniques for

    Human-Computer Interaction

    ANSHUL SHARMA

    Department of Electronics and Communication Engineering, MNNIT Allahabad

    Email ID: [email protected]

    Abstract- This paper presents some real-time

    video processing techniques for intelligent and

    efficient hand gesture recognition, with an aim

    of establishing a virtual interfacing platform for

    Human-Computer Interaction (HCI). The first

    step of the process is colour segmentation based

    skin detection, followed by area-based noise

    filtering. If a gesture is detected, the next step is

    to calculate a number of independent

    parameters of the available visual data and

    assign a distinct range of values of each

    parameter to a predefined set of different

    gestures. The final step is the hierarchical

    mapping of the obtained parameter values to

    recognise a particular gesture from the whole set.

    Deliberately, the mapping of gestures is not

    exhaustive, so as to prevent incorrect mapping

    (misinterpretation) of any random gesture not

    belonging to the predefined set. The applications

    of the same are inclusive of, but not limited to,

    Sign Language Recognition, robotics, computer

    gaming etc. Also, the concept may be extended,

    using the same parameters, to facial expression

    recognition techniques.

    1. INTRODUCTION

    Gestures and gesture recognition are terms

    increasingly encountered in discussions of human-

    computer interaction. The term includes character

    recognition, the recognition of proof readers

    symbols, shorthand, etc. Every physical action

    involves a gesture of some sort in order to be

    articulated. Furthermore, the nature of that gesture

    is generally an important component in establishing

    the quality of feel to the action. The general

    problem is quite challenging due a number of issues

    including the complicated nature of static and

    dynamic hand gestures, complex backgrounds, and

    occlusions. Attacking the problem in its generality

    requires elaborate algorithms requiring intensive

    computer resources. Due to real-time operational

    requirements, we are interested in a

    computationally efficient algorithm.

    Previous approaches to the hand gesture

    recognition techniques include the use of markers

    on various points on the hand, including fingertips.

    Calculation and observation of relative placement

    and orientation of these markers specifies a

    particular gesture. The inconvenience of placing

    markers on the users hand makes this an infeasible

    approach in practice. Another approach is to use

    sensor- fitted gloves to detect the orientation and

    other geometrical properties of the hand. The

    demerit of this approach is its cost ineffectiveness.

    The approach proposed in this text is quite user-

    friendly as it does not require any kind of markers

    or special gloves for its operation. Also, the

    memory requirements are low because the

    subsequent video frames are not stored in memory,

    they are just processed and overwritten. Obviously,

    it adds a new challenge to make the algorithm very

    fast and efficient, fulfilment of which is ensured by

    using low-complexity calculation techniques. For

    the ease of implementation, the proposed algorithm

    is based on three basic assumptions:

    1. The background should be dark. 2. The hand should always be at a constant

    distance from the camera.

    3. There should be a time gap of at least 200 ms between every two gestures.

  • CONFERENCE ON SIGNAL PROCESSING AND REAL TIME OPERATING SYSTEM (SPRTOS) MARCH 26-27 2011

    COPO306-2

    2. DATA EXTRACTION

    The concept of video processing is based upon

    data extraction from subsequent frames of the video.

    The video is accessed frame by frame and the

    following actions are performed on every

    subsequent frame: (Some frames may be skipped,

    without loss of information, because at an average

    frame rate of 30 fps, any two subsequent frames

    will have almost the same data.)

    2.1 Hand Block Detection:

    The first step of the gesture recognition process is

    to detect if there is a gesture at all. It is done by skin

    colour detection and area based detection of the

    hand block. Also, the gesture should only be taken

    into consideration if the whole hand is captured in

    the video frame. This is achieved by calculating the

    geometrical centroid of the hand block and

    imposing a range restriction on the coordinates of

    the centroid.

    2.2 Elliptical Approximation:

    The concept of elliptical approximation is the

    heart of the gesture recognition process discussed in

    this text. Once the hand block is detected, i.e. the

    presence of a gesture (valid or invalid) is ensured, it

    is approximated by an ellipse having the same

    second-moments as the hand block detected.

    Having obtained such a mathematical modelling,

    we can interpret a number of independent

    parameters from the geometrical properties of this

    ellipse. This concept is depicted in Figure 1.

    Figure 1: Elliptical approximation of hand block

    The Orientation of the hand object is simply the

    inclination of the major axis of the ellipse thus

    obtained, from the positive horizontal axis.

    Orientation is measured in a range of -90 to +90

    degrees, as depicted in Figure 2.

    Figure 2: Orientation of the hand block

    The Eccentricity of the ellipse is simply the ratio

    of semi-major axis (the distance between the two

    foci) to the major axis. The value of eccentricity of

    an ellipse lies between 0 and 1, both inclusive. An

    ellipse with zero eccentricity is actually a circle.

    The other extremity (eccentricity = 1) represents a

    straight line. Thus, the eccentricity is a measure of

    the narrowness of the ellipse. Higher be the

    eccentricity, narrower is the ellipse (and vice versa).

    2.3 Counting the number of fingers:

    A very fast approach to count the number of

    fingers in a gesture is discussed in this text, using

    the data extracted by the means of elliptical

    approximation as discussed already. With the vertex

    of the ellipse as centre and a radius equal to 0.35

    times the length of major axis, a circle is drawn.

    The number of times the periphery of this circle

    intersects any part of the hand block gives a direct

    measure of the number of fingers in the gesture.

    Figure 3 explains the concept. Clearly, if the circle

    intersects the hand block at N places, then (N-1) is

    the number of fingers in the gesture. As far as the

    selection of the radius is concerned, 0.35 times the

    major axis length was experimentally found to be

  • CONFERENCE ON SIGNAL PROCESSING AND REAL TIME OPERATING SYSTEM (SPRTOS) MARCH 26-27 2011

    COPO306-3

    the most optimum. A larger scaling factor (greater

    than 0.4) entirely covers the smaller fingers and

    hence, the presence of thumb and the little finger in

    a gesture cant be detected. On the other hand, a

    small scaling factor (less than 0.3) is unable to

    distinguish between two fingers separated by a

    small gap, because the gap between the fingers is

    lesser near the palm.

    Figure 3: Counting the number of fingers

    2.3 Perimeter Estimation:

    Another distinguishing characteristic of a gesture

    is the length of the periphery of the hand object.

    Perimeter is calculated by simply counting the

    number of pixels on the boundary of the hand

    object.

    Figure 4(a)

    Figure 4(b)

    Figure 4: Gestures having equal Areas but unequal Perimeters

    Its significance can be appreciated by

    acknowledging the fact that two gestures with same

    area can have different perimeters. The hand

    objects in Figure 4(a) and Figure 4(b) cover

    approximately the same area, but the perimeter of

    the hand object in Figure 4(a) is larger than that of

    the hand object in Figure 4(b).

    3. GESTURE RECOGNITION PROCESS

    Having calculated the various parameters of the

    hand block by mathematical modelling using

    elliptical approximation, we now assign a unique

    set of values of all these parameters to every gesture,

    to be able to distinguish each one uniquely. Instead

    of calculating all the parameters for every gesture, it

    is efficient to categorize the gestures into different

    hierarchical levels, so that only relevant properties

    of every gesture are calculated, to make the

    algorithm time-efficient. Statistically speaking, the

    most optimum order is achieved by adopting the

    following hierarchical recognition process:

    1. Eccentricity. The first step is to mathematically

    model the gesture by elliptical approximation and

    calculate the eccentricity of the ellipse thus

    obtained. If the eccentricity is too low (less than

    0.5), it shows that the ellipse is wide enough

    (approaching a circular shape), as is the case with

    the gestures shown in Figure 5.

  • CONFERENCE ON SIGNAL PROCESSING AND REAL TIME OPERATING SYSTEM (SPRTOS) MARCH 26-27 2011

    COPO306-4

    (a)

    (b)

    Figure 5: Gestures with Eccentricity < 0.5

    On the other hand, if the eccentricity is too high

    (greater than 0.95), it shows that the ellipse is very

    narrow (almost a straight line), which is the case

    with the gestures shown in Figure 6. In both these

    cases, there is no need to count the number of

    fingers, as it is obviously zero.

    2. Number of Fingers. If the value of Eccentricity

    lies between 0.5 and 0.95 (both exclusive), the

    algorithm to calculate the number of fingers is

    applied to the gesture frame.

    3. Perimeter Estimation. Perimeter estimation is

    necessary as a distinguishing parameter only in a

    few cases, when the number of fingers is less than

    three. This is obvious from Figure 4, which shows

    one case in which perimeter estimation is required

    as a distinguishing feature. Another similar case is

    complimentary to these gestures, with negative

    Orientation.

    Figure 6: Gestures with Eccentricity > 0.95

    4. Orientation. Orientation calculation is necessary

    for every gesture, irrespective of the values of the

    other parameters associated with it. This is because

    the proposed gesture set is so made that every

    gesture has a complimentary gesture whose all the

    properties are the same as the former, with the only

    difference being that its Orientation is the negative

    of that of former. Since Orientation calculation is a

    must for every gesture, this step may be executed at

    any level, without any effect on the complexity of

    the algorithm. Keeping it at the end of the process is

    just one of the several equally efficient options.

    4. THE ENTIRE GESTURE SET

    The algorithm and the parameters already

    discussed altogether can uniquely recognise a

    gesture out of a closed set of 32 distinct gestures.

    The complete gesture set is shown in Figure 7.

    5. APPLICATIONS

    Real-time gesture recognition is an emerging and

    challenging field of Computer Vision. The main

    applications of the same include, but are not limited

    to:

  • CONFERENCE ON SIGNAL PROCESSING AND REAL TIME OPERATING SYSTEM (SPRTOS) MARCH 26-27 2011

    COPO306-5

    Figure 7: The Entire Gesture Set

    Human- Computer Interaction (HCI).

    Wireless robotic control mechanism.

    Virtual Keyboard Interfacing Platform.

    Security systems.

    6. SCOPE FOR FUTURE WORK

    The proposed gesture recognition techniques can

    prove to be an integral basis for research work in

    certain emerging fields, viz. Computer Vision, Real

    Time Signal Processing, HCI, etc. Some major

    fields that provide a challenging scope for the

    future work of this text are discussed below:

    Motion Gestures: So far we have discussed only the recognition of static gestures. The above

    mentioned techniques, coupled with a concept

    of motion recognition, can be used for

    recognition of motion gestures.

    No Background Limitation: The above mentioned techniques are based on the

    assumption that the background is always dark.

    However, using dynamic colour segmentation

    schemes, the same techniques can be applied

    without this limitation also.

    Double-Handed Gestures: All the same concepts can be easily extended to gestures

    involving both the hands at once. Doing so will

    increase the size of the gesture set by about 5

    times (with a substantial increase in complexity,

    though).

    Three- dimensional Gestures: The proposed techniques deal with the recognition of only

    two-dimensional gestures (hence the

    assumption that the hand should always be at

    constant distance from the camera, as discussed

    earlier). However, if we use an additional

    camera, orthogonal to the first one, we can

    obtain the information about the distance of the

    hand from the camera. This can be helpful for

    efficient recognition of three-dimensional

    gestures.

    Facial Expression Recognition: The concept of mathematical modelling of gestures using

    Elliptical Approximation can be quite efficient

    for the recognition of facial expressions.

    Artificial Intelligence: Real-time Gesture Recognition, coupled with a concept of Neural

    Networks, may prove to be an integral

    milestone in the field of Artificial Intelligence.

    7. REFERENCES

    [1]. Juan P. Wachs, Helman Stern, Yael Edan: Cluster

    Labeling and Parameter Estimation for the Automated Setup

    of a Hand-Gesture Recognition System, IEEE Transactions

    on Systems, Man and CyberneticsPart A: Systems and

    Humans, Vol. 35, No. 6, November 2005.

    [2]. Hyung-Ji Lee, Jae-Ho Chung: Hand Gesture Recognition

    Using Orientation Histogram, 1999 IEEE Tencon.

    [3].Rafael C. Gonzalez, Richard Eugene Woods: Digital

    Image Processing, Third Edition, Prentice Hall Publications,

    2008.

  • CONFERENCE ON SIGNAL PROCESSING AND REAL TIME OPERATING SYSTEM (SPRTOS) MARCH 26-27 2011

    COS0102-1

    Modelling Neuron for Biomedical Applications: A Review

    1Taslima Ahmed , 2 Dr Jiten Ch Dutta

    1Dept. of Electronics and Instrumentation Engg. IIMT College of Engineering, Greater Noida, Knowledge Park-III, UP-201306, India

    2Dept of ECE, Tezpur University , Napaam Post, Tezpur, Assam 784 028, India

    [email protected]

    Abstract: Modelling of neuron including the action of synapse has played an important role in the

    field of biomedical engineering and neurology for simulation of receptor function and electrical

    activity of the postsynaptic neuron. In this paper, we review some literatures concerning the

    development of different neuron models giving special emphasis on Dutta and Roy models.

    Keywords Neuron, Synapse, MOSFET, Postsynaptic membrane.

    Over past few years, many electronic circuits

    have been developed to reproduce the behaviour of

    nerve axons [1]-[5]. A very good account of this

    type of modelling is reviewed by Harmon et al [6]

    and Lewis [7]. But among these models,

    neuroscientists have, so far, utilized Hodgkin-

    Huxley (H-H) model as a circuit analog of the

    axonal membrane. In this model, the capacitance of

    the lipid bilayer of postsynaptic membrane is

    represented by CM and is found to be constant and

    the membrane resistance is determined in terms of

    three parallel conductances gNa, gk, and g0 as shown

    in Fig.1. The conductances gNa, gK, and go represent

    the membrane permeability of Sodium, Potassium

    and other ions respectively. ENa, and EK are

    respectively the chemical potentials of Sodium and

    Potassium i.e., Nernstian membrane potential for

    Sodium and Potassium. EO is the resting potential.

    The gk and gNa conductances are found to be time

    and voltage dependent.

    Fig.1: H-H model

    The total current in this model is given by:

    I = Im+Io-INa+IK (1)

    If Vm be the postsynaptic membrane potential

    established by the ionic and capacitive membrane

    current then

    I = C(dVm/dt)+gO(VmEO)gNa(VmENa)+gK

    (VmEK) (2)

    The equations (1) and (2) are called H-H equations

    which are simple and capable of explaining the

    activity of neuron with the help of variable

    permeability of membrane for different ions, e.g.,

    sodium, potassium and other ions. But this model

    has not explained the function of synapses on which

    the variable permeability of postsynaptic membrane

    arises.Refering to the biological activities of neuron,

    the primary mode of communication between two

    neurons is a biochemical process that occurs at

    synapse. Synapse is essentially a junction called

    synaptic cleft between two neurons namely

    presynaptic and postsynaptic neurons. Signal from

    presynaptic neuron to postsynaptic neuron is

    transmitted through neurotransmitters released by

    presynaptic neuron terminals in to the synaptic cleft.

    Neurotransmitters diffuse through the cleft and then

    bind with the specific receptor sites of the

    membrane of postsynaptic neuron. This binding

    mechanism initiates the opening of transmitter

  • CONFERENCE ON SIGNAL PROCESSING AND REAL TIME OPERATING SYSTEM (SPRTOS) MARCH 26-27 2011

    COS0102-2

    gated ion channels resulting in to flow of ions into

    the cell or out of the post synaptic cell.

    The membrane of post synaptic neuron has two

    types of ion channels excitatory and inhibitory.

    The excitatory channels are those which are specific

    to sodium ions and inhibitory channels are those

    which are specific to Chloride ions. The flow of

    Sodium ions into the cell causes a membrane

    potential called excitatory postsynaptic membrane

    potential (EPSP) whereas the flow of Chloride ions

    causes an inhibitory postsynaptic membrane

    potential (IPSP).The electrical mechanism of

    synapse is shown in Fig 2. When an action potential

    from the presynaptic neuron arrives at its terminals

    connecting the cleft, neurotransmitters are released

    into the cleft which diffuse through the cleft and

    bind with the receptor sites of the postsynaptic

    membrane. This binding mechanism opens the ion

    channels situated at the membrane surface and ions

    move into or out of the membrane. If the synapse is

    excitatory, Sodium ions flow into the cell resulting

    into positive current. As a result the membrane

    depolarizes. If sufficient number of Sodium

    channels open, then membrane potential will be

    greater than the threshold VT of the neuron and

    initiates an action potential. If the synapse is

    inhibitory, Chloride ions move into the cell,

    resulting into negative current. As a result the

    membrane hyperpolarizes. If the numbers of

    opening of Chloride channels are sufficiently large

    then membrane potential will be able to initiate an

    action potential in negative direction. The

    presynaptic equivalent circuit is shown in Fig 3(a),

    where I is the total current from ionic channels of

    all synapses and E1 ,E2, ., EM represent the

    chemical potentials of each corresponding ions. For

    example, EM may be ENa or may be ECl. The total

    current I will stimulate the postsynaptic neuron to

    initiate an action potential [8]. Fig 3(b) shows the

    equivalent circuit of a synapse which is developed

    by adding H-H equivalent circuit with the

    presynaptic circuit shown in Fig 3(a) [9].The

    postsynaptic membrane consists of a lipid bilayer

    and transmembrane protein ion channels. Some ion

    channels such as sodium, chloride etc. are

    controlled by the neurotransmitters that bind with

    the receptor sites, i.e. the

    Fig.2: Electrical mechanism of synapse

    Fig.3(a): Equivalent circuit of a presynaptic neuron

    Fig. 3(b): Electrical equivalent circuit of synapse

    amount of ionic current is dependent upon the

    activity of the transmitter-receptor binding. In

    simplest case, the binding reaction may be

    represented as

  • CONFERENCE ON SIGNAL PROCESSING AND REAL TIME OPERATING SYSTEM (SPRTOS) MARCH 26-27 2011

    COS0102-3

    Neuro-transmitter+Receptor(Closed)1

    2

    K

    K

    Neuro-

    transmitter Receptor(Open) (3)

    Where K1 and K2 are the forward and backward

    rate constants respectively. The transmitter gated

    channels, therefore, have variable conductance

    dependence on the binding activity of transmitters.

    Dutta and Roy therefore have modeled transmitter

    gated ion channels by MOSFET, because MOSFET

    functions as a voltage controlled conductance in its

    linear region [10]. In this model they have

    considered gate voltage as a time dependent voltage

    given by

    Vg(t)=V0[(1exp(-k1t)+exp(-k2t)U(t-tm)]

    (4)

    Where K1 and K2 are time constants analogous to

    the rate constants of equation (3), U(t-tm) is the

    Heaviside function andVo is a voltage proportional

    to the maximum attainable conductance, when all

    the transmitter-gated channels for a specific ion are

    open. Based on this, they have developed a

    biologically motivated model as shown in Fig.4 (a).

    Their circuit models both for excitatory and

    inhibitory synapses are shown in Fig.4(b) and 4(c)

    respectively. In both these models they have

    divided the postsynaptic membrane into three

    patches to represent spatial summation of the

    sodium current and chlorine current controlled by

    respectively sodium and chloride conductances [10].

    Fig.4(a): Biologically motivated model of

    postsynaptic membrane.

    Fig.4(b): Circuit model for excitatory action of

    synapse

    Fig.4(c): Circuit model for inhibitory action of

    synapse

    The simulation results from this model [Fig.5]

    indicate that this model can be used in neuro

    bioengineering area for simulation of

    neurotransmitter-receptor binding activity and

    electrical activity of the postsynaptic neuron.

  • CONFERENCE ON SIGNAL PROCESSING AND REAL TIME OPERATING SYSTEM (SPRTOS) MARCH 26-27 2011

    COS0102-4

    Fig.5: Simulation results of excitatory and

    inhibitory actions of postsynaptic Membrane. Top

    waveform represents the EPSP and bottom

    waveform represents the IPSP.

    In an another model, they have used ion sensitive

    field effect transistor (ISFET) as circuit analog and

    incorporated into the famous Hodgkin-Huxley (H-H)

    model of neuron to substitute the variable Na+ and

    Cl- conductances, the details of which may be

    obtained in reference [11]. They have mentioned

    that such model has additional advantages over

    MOSFET based model. The advantages that they

    have high lighted are: (i) Measurement of different ions that diffuse through the post synaptic membrane and hence pH (ii) measurement of neurotransmitters diffused through the synaptic cleft by converting the ISFET into neurotransmitter sensitive enzyme modified FET (ENFET). This model, according to them, may become a useful research unit in neurology for biotelemetry applications. The second advantage that ENFET can also be used as circuit analog, in an effort, they have modeled it using ENFET sensitive to acetylcholine neurotransmitter for simulation of acetylcholine gated ion channels of the post synaptic membrane at the synaptic cleft[12].

    MOSFET and ISFET based electrical models both

    for excitatory and inhibitory actions of neurons

    have been reviewed. It is concluded that ISFET

    based models are more biologically motivated as

    these are compatible with biological medium and

    there is possibility of measurement of

    neurotransmitters diffused through the synaptic

    cleft by converting the ISFET into neurotransmitter

    sensitive ENFET. These biologically motivated

    models may become useful research and teaching

    units in biomedical area in general and neurology in

    particular.

    REFERENCES [1] Hodgkin, A, L and Huxley, A. F., A

    quantitative description of membrane current and its application to conduction and excitation in nerve, J. Physiol, 117. 500-544(1952)

    [2] Hodgkin, A. L., Ionic movements and electrical activity in giant nerve fibers, Proceedings of the Royal Society of London. Series B, Biological Sciences, Vol. 148, 1-38(1957)

    [3] Fitzhugh, R., Threshold and plateaus in the Hodgkin-Huxley nerve equations, J. Gen. Physiology, 43, 867-(1960)

    [4] Johnson and Hanna, Membrane model: A single transistor analog of excitable membrane, J. Theoret. Bio, 22, 401-411(1969)

    [5] E.R. Lewis, Neuroelectric potentials derived from an extended version of the Hodgkin and Huxley model, J. Theor. Biol. Vol.10,125-158, 1965

    [6] L.D. Harmon and E.R.Lewis, Neural modelling, Physiol. Rev. , Vol, 48, 513-591, 1966

    [7] E.R.Lewis, Using electronic circuits to model simple neuroelectric interactions, Proc.IEEE,vol 56, 931-949, June 1968.

    [8 ] Xiao-lin Zhang, A Mathematical Model of a Neuron with Synapses

    based on Physiology, Nature Proceedings,

    npre.2008.1703.1. March

    2008.

    [9] Soumik Roy, Jiten Ch Dutta, Shikhamoni

    Phukan, Integrate-and-

    Fire based Circuit model for simulation of

    excitatory and inhibitory

    synapses, Canadian Journal on Biomedical

    Engineering & Technology

    Vol. 1, No. 2 March 2010, 49-51. [10] Jiten Ch Dutta and Soumik Roy

    Biologically motivated Circuit model for simulation of excitatory and inhibitory synapses, Canadian Journal on Biomedical Engineering & Technology Vol. 1, No. 2 June 2010, 49-51.

    [11] Jiten Ch Dutta and Soumik Roy, An Electronic Circuit Model for simulation of Synaptic Communication: The NEUROISFET for Wireless Biotelemetry, accepted for publication in the IEEE conf. on wireless communication, 24-25 Feb, 2011, BITS, MESRA

    [12] Jiten Ch Dutta and Soumik Roy Biologically inspired Circuit model for simulation of Acetylcholine gated ion channels of the Postsynaptic membrane at synaptic cleft,

  • CONFERENCE ON SIGNAL PROCESSING AND REAL TIME OPERATING SYSTEM (SPRTOS) MARCH 26-27

    2011

    SIP0109-1

    Electricity Generation by People Walk through

    Piezoelectric Shoe: An Analysis

    1.Dr. Monika Jain,

    2.Ms. Usha Tiwari,

    3.Mohit Gupta,

    4.Magandeep singh Bedi

    1.Member IEEE, IETE, Professor-Dept of Electronics & Instrumentation Engg,

    Galgotias College of Engineering & Technology,Greater Noida,UP, INDIA 2.

    Assistant Professor-Dept of Electronics & Instrumentation

    Galgotias College of Engineering & Technology,Greater Noida,UP, INDIA .3&4.

    B.Tech, 4th

    year student Dept of Electronics & Instrumentation

    Galgotias College of Engineering & Technology,Greater Noida,UP, INDIA [email protected]

    [email protected].

    [email protected]

    [email protected]

    Abstract In todays, high crisis of

    electrical power, there has been an

    increasing demand for low-power and

    portable-energy sources due to the

    development and mass consumption of

    portable electronic devices.

    Furthermore, the portable-energy

    sources must be associated with

    competitive market price,

    environmental issues and other imposed

    regulations. These tremendous demands

    support lots of research in the area of

    portable-energy generation methods. In

    this scope, piezoelectric materials has

    always been chosen as an attractive

    choice for energy generation and

    storage. In this paper, different

    techniques are being explored and

    analysed to generate electricity by usage

    of piezoelectric crystal. In-depth study

    and analysis to describes the use of

    piezoelectric polymers in order to utilize

    and the best optimisation of energy

    from people-walk and the fabrication of

    a smart shoe, capable of generating and

    accumulating the energy has peen

    presented.

    Keywords Energy harvesting, PZT,

    uninterrupted power supplies.

    I. INTRODUCTION

    Piezoelectric generators are based on

    piezoelectric effect i.e. the ability of

    certain materials to create electrical

    potential when responding to mechanical

    changes. In real time application, when

    compressed or expanded or otherwise

    changing shape a piezoelectric material

    will output certain voltage. This effect is

    also possible in reverse in the sense that

    putting a charge through the material will

    result in it changing shape or undergoing

    some mechanical stress. These materials

    are useful in a variety of ways. Certain

    piezoelectric materials can handle high

    voltage extremely well and are useful in

    transformers and other electrical

    components. Piezoelectric crystals are

    boon of sensor technology field as it might

    be possible to make motors, reduce

    vibrations in sensitive environments, used

    as an energy collector and in many more

    applications. In todays power crisis world,

    one of the most interesting area is energy

    collection and generation. In this paper, a

    cheap and smart however a reliable

    mechanism to generate energy capable

    enough to charge our phone, MP3 players

    has been explored and analysed. An

    interesting methodology of power

  • CONFERENCE ON SIGNAL PROCESSING AND REAL TIME OPERATING SYSTEM (SPRTOS) MARCH 26-27

    2011

    SIP0109-2

    generation through the walking steps of

    human being is reviewed and presented

    here. The sole of shoe could be constructed

    of piezoelectric materials and every step a

    person took would begin to generate

    electricity. This smart mechanism of

    generation of electricity through shoe sole

    could then be stored in a battery or used

    immediately in personal electronics

    devices.

    II. LITERATURE REVIEW

    The most common methodology of

    shoe power generators include dielectric

    elastomers [1] and piezoelectric ceramics

    [2,3]. The elastomer demonstrated

    significant power output but it required a

    large bias (2 kV) and the heavy

    construction is likely to negatively affect

    the user experience. The power harvesting

    shoe reported in [2] and [3] uses

    piezoelectric ceramic bi-morphs for power

    harvesting. As piezoelectric materials were

    employed, no bias voltage was needed.

    However, a complex PZT/metal bi-morph

    was required and the power output after

    dc/dc conversion and regulation was low

    (

  • CONFERENCE ON SIGNAL PROCESSING AND REAL TIME OPERATING SYSTEM (SPRTOS) MARCH 26-27

    2011

    SIP0109-3

    There are two types of piezoelectric

    signals that can be used for technological

    applications: the direct piezoelectric effect

    that describes the ability of a given

    material to transform mechanical strain

    into electrical signals and the converse

    effect, which is the ability to convert an

    applied electrical solicitation into

    mechanical energy. The direct

    piezoelectric effect is more suitable for

    sensor applications, whereas the converse

    piezoelectric effect is most of the times

    required for actuator applications[12].

    High-performance films, prepared by

    researchers [14-15] is also explored. In this

    the electromechanical properties of the

    film were improved by a treatment that

    consists of pressing, stretching, and poling

    at a high temperature [14].

    III. CONCLUSION

    In this paper, an analysis for Electricity-

    Genration for low power devices is done.

    Different methodologies for generation of

    electricity is reviewed and presented. We

    analysed that some of the methodologies

    are not feasible due to too much circuitry

    in real time portable charging and some

    are feasible but they are on an analysis

    stage. We have found that piezoelectric

    generators implanted in shoe can provide

    a great achievement if collaboratively an

    effort is made to bring a commercial

    battery charger for low power house

    devices, just by utilization of walking steps

    of a person.

    REFERENCES

    [1] Roy Kornbluh, Power from plastic:

    how electroactive polymer artificial

    muscles will improve portable ower

    generation in the 21st century military,

    Presentation [Online],

    Available:http://www.dtic.mil/ndia/2003tri

    service/korn.ppt

    [2] John Kymisis, et.al., Parasitic power

    harvesting in shoes in Proc. of the 2nd

    IEEE Int. Conf. On Wearable Computing,

    Pittsburgh. PA, pp. 132-139, 19-20 Oct.

    1998.

    [3] S. Shenck and J. Paradiso, Energy

    scavenging with shoe-mounted

    piezoelectrics, IEEE Micro, Vol. 21, pp.

    30-42, May-June, 2001.

    [4] P. Miao, et.al., Micro-Machined

    Variable Capacitors for Power

    Generation, in Proc. Electrostatics

    Edinburgh, UK, 23-27 Mar. 2003.

    [5] Mitcheson, P.D.; Green, T.C.;

    Yeatman, E.M.; Holmes, A.S.,

    "Architectures for vibration-driven

    micropower generators," Journal of

    Microelectromechanical Systems, vol.13,

    no.3, pp. 429-440, June 2004.

    [6] Ville Kaajakari, Practical MEMS,

    Small Gear Publishing, 2009.

    [7] M.Duffy & D.Carroll,

    Electromagnetic generators for power

    harvesting 35th AM^ IEEE Power

    Electronics Specialists Conference

    Aachen, Germany, 2004 ;pp. 2075-2081

    [8]M. El-hami, P. Glynne-Jones, M.

    White, M. Hill, S. Beeby, E. James, D.

    Brown, and N. Ross, Design and

    fabrication of a new vibration-based

    electromechanical power generator,

    Sens. Actuators A, Phys., vol. 92, no. 13,

    pp. 335342, Aug. 2001.

  • CONFERENCE ON SIGNAL PROCESSING AND REAL TIME OPERATING SYSTEM (SPRTOS) MARCH 26-27

    2011

    SIP0109-4

    [9] M. Miyazaki, H. Tanaka, G. Ono, T.

    Nagano, N. Ohkubo, T. Kawahara, and K.

    Yano, Electric-energy generation using

    variablecapacitive resonator for power-

    free LSI, in Proc. ISLPED, 2003, pp.

    193198.

    [10] C. Keawboonchuay and T. G. Engel,

    Maximum power generation in a

    piezoelectric pulse generator, IEEE

    Trans. Plasma Sci., vol. 31, no. 1, pp. 123

    128, Feb. 2003.

    [11] J. Yang, Z. Chen, and Y. Hu, An

    exact analysis of a rectangular plate

    piezoelectric generator, IEEE Trans.

    Ultrason., Ferroelectr., Freq. Control, vol.

    54, no. 1, pp. 190195, Jan. 2007.

    [12] T. Sterken, P. Fiorini, K. Baert, R.

    Puers, and G. Borghs, Anelectret-based

    electrostatic micro-generator, in Proc.

    Transducers,2003, pp. 12911294.

    [14] V. Sencadas, R. Gregorio Filho, and

    S. Lanceros-Mendez, Processing and

    characterization of a novel nonporous

    poly(vinilidene fluoride) films in the

    phase, J. Non-Cryst. Solids, vol. 352, no.

    21/22, pp. 22262229, Jul. 2006.

    [15] S. Lanceros-Mendez, V. Sencadas,

    and R. Gregorio Filho, A new

    electroactive beta PVDF and method for

    preparing it, Patent PT103 318, Jul. 19,

    2006.

  • CONFERENCE ON SIGNAL PROCESSING AND REAL TIME OPERATING SYSTEM (SPRTOS) MARCH 26-27 2011

    SIP0304-1

    EYE BASED CURSOR MOVEMENT USING EEG IN

    BRAIN COMPUTER INTERFACE Tariq S Khan

    #, Mudassir Ali

    #, Omar Farooq

    #, Yusuf U Khan

    *,

    #Department of Electronics Engineering, Zakir Husain College of Engineering & Technology

    *Department of Electrical Engineering, Zakir Husain College of Engineering & Technology

    Aligarh Muslim University, Aligarh

    Abstract The aim of this study is to

    detect eye movement (left to right) from

    Electroencephalograph (EEG) signal.

    Four electrodes of EEG in the frontal

    area were used. The statistical features

    were extracted from the four channels of

    frontal channel. These features were then

    fed into a classifier based on the linear

    discriminator function. The most

    prominent features for the classification

    of left and right movements were

    identified. These features were then

    interfaced with computer so that cursor

    movement can be controlled. Electrodes

    are placed along the scalp following the

    10-20 International System of Electrode

    Placement. Recorded data was filtered,

    windowed and analysed in order to

    extract features. Four different classifiers

    were used. Best results were found in

    support vector machine (SVM) and linear

    classifiers each of which gave the average

    accuracy of 90%.

    Keywords: BCI, Eye movement, EEG.

    I. INTRODUCTION

    A brain-computer interface (BCI)

    provides an alternative communication

    channel between the human brain and a

    computer by using pattern recognition

    methods to convert brain waves into control

    signals. Patients who suffer from severe

    motor impairments (severe cerebral palsy,

    head trauma and spinal injuries) may use

    such a BCI system as an alternative form of

    communication by mental activity [1]. Using

    improved measurement devices, computer

    power, and software, multidisciplinary

    research teams in medicine,

    psychophysiology, medical engineering, and

    information technology are investigating and

    realizing new noninvasive methods to

    monitor and even control human physical

    functions.

    In a bigger picture there can be devices

    that would allow severely disabled people to

    function independently. For a quadriplegic,

    something as basic as controlling a computer

    cursor via mental commands would

    represent a revolutionary improvement in

    quality of life. With an EEG or implant in

    place, the subject would visualize closing his

    or her eyes or moving eyes from left to right

    and vice versa [2]. The software can learn

    eye movement through training, using

    repeated trials. Subsequently, the classifier

    may be used to instruct the closure/opening

    of eye. A similar method is used to

    manipulate a computer cursor, with the

    subject thinking about forward, left, right

    and back movements of the cursor [3]. With

    enough practice, users can gain enough

    control over a cursor to draw a circle, access

    computer programs and control a television.

    It could theoretically be expanded to allow

  • CONFERENCE ON SIGNAL PROCESSING AND REAL TIME OPERATING SYSTEM (SPRTOS) MARCH 26-27

    2011

    SIP0304-2

    users to "type" with their thoughts. This can

    be achieved by controlling cursor movement

    on a computer screen through EEG signals

    from brain, specifically, generated due to

    eye movement. The signals can be analysed

    by different methods.

    Traditional analysis methods, such as the

    Fourier Transform and autoregressive

    modelling are not suitable for non-stationary

    signals. Recently, wavelets have been used

    in numerous applications for a variety of

    purposes in various fields. It is a logical way

    to represent and analyse a non-stationary

    signal with variable sized region windows

    and to provide local information. In the

    Fourier Transform (FT), the time

    information is lost and in short Term Fourier

    Transform (STFT) there is limited time

    frequency resolution. Even though basic

    filters can be used for decomposition of

    desired bands, ideal filters are never realised

    in practice, which results in aliasing effects.

    However, wavelet analysis enables perfect

    decomposition of the desired bands, which

    helps us to obtain better features [4].

    In this paper different features are used

    for training the classifier for eye movement

    in left and right directions. A time-frequency

    analysis was applied to the EEG signals

    from different channels, to determine

    combination of features and channels that

    yielded the best classification performance.

    II. BACKGROUND RESEARCH

    EEG waves are created by the firing of

    neurons in the brain and were first measured

    by Vladimir Pravdich-Neminsky who

    measured the electrical activity in the brains

    of dogs in 1912, although the term he used

    was electrocerebrogram. Ten years later

    Hans Berger became the first to measure

    EEG waves in humans, in addition to giving

    them their modern name, began what would

    become intense research in utilizing these

    electrical measurements in the fields of

    neuroscience and psychology.

    The term Brain-Computer Interface first

    appeared in scientific literature in the 1970's,

    though the idea of hooking up the mind to

    computers was nothing new [5]. Currently,

    the systems are open loop and responds to

    users thoughts only. The closed loop

    systems are aimed to be developed that can

    give feedback to user as well.

    In order to meet the requirements of the

    growing technology expansion, some kind of

    standardization was required not only for the

    guidance of future researchers but also for

    the validation and checking of new

    developments with other systems, thus a

    general purpose system was developed

    called BCI2000 which made analysis of

    brain siganl recording easy by defining the

    output formats and operating protocols to

    facilitate the researchers in developing any

    type of application. This made it easier to

    extract specific features of brain activity and

    translate them into device control signals

    [7]..

    III. OUR METHODOLOGY

    The procedure in this study was to initially

    acquire EEG data. The stored data was then

    pre-processed to remove artifacts.

    Subsquently features were extracted in the

    clean EEG and used for classification. Thus

    methodology is shown in Fig. 1.

    Fig. 1: Block diagram for feature extraction and device

    control of eye movement

    Data acquisition

    Data Processing

    Feature Extraction

    Classification

    Device/Application Control

  • CONFERENCE ON SIGNAL PROCESSING AND REAL TIME OPERATING SYSTEM (SPRTOS) MARCH 26-27

    2011

    SIP0304-3

    A.Experimental Setup and Data Acquisition

    The subject was seated on wooden armchair

    and legs were rested on wooden footrest

    (wooden items should be used so as to

    reduce interference) with eyes closed. The

    subject was instructed to avoid speaking and

    to avoid body movement in order to ensure

    relaxed body. EEG data were recorded using

    a Brain Tech clarityTM

    system [9] with the

    electrodes positioned according to the

    standard 10-20 system in the biomedical

    Signal Processing lab, AMU Aligarh.

    To ensure the same rate of eye movement in

    both directions, a ball was shown on the

    screen and the subject was asked to visually

    follow the ball. The movement of ball was

    set to 60 pixels per second. A series of trials

    were recorded.

    The subject was instructed to open eyes

    slowly and then to follow the movement of

    the ball in the program on prompt from the

    experimenter. Movement of eyes was

    recorded for two different directions i.e. left

    to right and right to left. Block diagram of

    experimental procedure is shown in fig. 2.

    Fig. 2 Sequence followed during experimental recording

    B. Data Processing

    26 channels of EEG were recorded. Since

    only frontal lobe is mainly involved in eye

    movement, only those channels that are

    associated with the frontal lobe i.e. FP1-F3,

    FP1-F7, FP2-F4, FP2-F8 were analysed. The

    signal values associated with these signals

    were extracted in ASCII form using

    BrainTech software. EEG of the frontal lobe

    channels for subject 1 is illustrated in Fig. 3.

    Fig. 3: Plot of channels associated with frontal lobe

    50 Hz power supply often causes

    interference in the EEG recording. Fig. 4

    shows a plot of PSD on the EEG record of

    FP1-F3 channel. To eliminate these spikes

    signal was passed through Infinite Impulse

    Response (IIR) notch filter before analysis.

    Fig. 4: Power Spectral Density of FP1F3

    IIR second order notch filter with the

    quality factor (or Q factor) of 3.91 was used

    to remove the undesired frequency

    components.

    Signal after removing the artifacts of 4

    channels stacked over one another is shown

    in Fig. 5.

    Relax

    Left to

    right

    movement

    Relax

    Right to

    left

    movement

    50 100 150 200 250 300 350 400 450 500

    0

    500

    1000

    1500

    No. of Samples

    Am

    pli

    tud

    e

    Frontal lobe channels

    fp1f3

    fp1f7

    fp2f4

    fp2f7

    0 20 40 60 80 100 120 140-40

    -30

    -20

    -10

    0

    10

    20

    30

    Frequency (Hz)

    Po

    wer

    (dB

    )

    PSD before & after Passing Through Notch

    PSD of FP2F8

    PSD after passing

    through Notch Filter

  • CONFERENCE ON SIGNAL PROCESSING AND REAL TIME OPERATING SYSTEM (SPRTOS) MARCH 26-27

    2011

    SIP0304-4

    Fig. 5: Signal Plot of filtered Frontal lobe associated

    channels

    EEG by nature is non stationary signal. So it

    was fragmented into frames so that it can be

    assumed stationary for small segment. EEG

    data is divided into frames of 1s duration i.e.

    frame size of 256 samples.

    C. Feature extraction

    Feature extraction is the process of

    discarding the irrelevant information to the

    possible extent and representing relevant data

    in a compact and meaningful form. Two eye

    movements were recorded: right to left

    (RTL), left to right (LTR).Standard statistical

    parameters such as mean, variance,

    skewness, cross-correlation were calculated

    for all the channels in each movement type.

    D. Classification

    Following classifiers were used to classify

    the two eye movements:

    SVM: It is non-probabilistic binary linear

    classifier.

    Linear: Fits a multivariate normal density to

    each group, with a pooled estimate of

    covariance.

    Diaglinear: Similar to 'linear', but with a

    diagonal covariance matrix estimate (naive

    Bayes classifiers).

    Quadratic: Fits multivariate normal densities

    with covariance estimates stratified by

    group.

    E. Cursor Control

    A program was written which controls the

    cursor movement according to instruction

    given. This program will be calibrated

    according the instructions given i.e. the

    cursor movement will be invoked instead of

    mouse movement as the instruction same as

    that of the mouse movement. This

    instruction will then be interfaced with the

    eye movement which will then control the

    movement of cursor [9].

    IV. RESULTS AND DISCUSSIONS

    For each frame of EEG, four features were

    calculated namely, variance, mean, skewness

    and cross correlation. The seperabrability

    provided by each feature was individually

    tested. The best three features were

    subsequently used as an input to the

    classifier. Four classifiers were used in this

    work. The classifiers results are illustrated in

    Table 1.For each movement of LTR and

    RTL 20 seconds (20 frames) of data were

    collected. From these 20 frames 15 frames

    were used for training and rest 5 are used for

    testing for both movements.

    Table 1: Percentage accuracy of classification for eye

    movements

    Classifier RTL LTR

    SVM 80 100

    Linear 80 100

    Quad 60 40

    Diaglinear 80 60

    From the observations in Table1 it can be

    seen that linear or SVM classifier gives the

    best possible results with high classification

    percentage accuracy for both eye

    movements.

    0 100 200 300 400 500-500

    0

    500

    1000

    1500

    2000Signal after passing through Notch filter

    fp1f3

    fp1f7

    fp2f4

    fp2f8

  • CONFERENCE ON SIGNAL PROCESSING AND REAL TIME OPERATING SYSTEM (SPRTOS) MARCH 26-27

    2011

    SIP0304-5

    Fig. 6: Plot of Classifier in Signal Space

    A linear classifier classifying both eye

    movements is shown in Fig. 6.

    Fig. 7: Variance Plot of FP2-F4

    From Fig. 6 which shows the variance for the

    channel FP2-F4 clearly shows that the

    variance of LTR is greater than RTL for

    most of the time. Variance basically shows

    the concentration of probability density

    function about the mean.

    V. CONCLUSIONS

    EEG data was investigated for two eye

    movements using a 4 channel setup on three

    subjects. Features were extracted from the

    variance for both the movements. A linear

    classifier was used to classify between the

    two eye movements. These algorithms can

    provide high classification accuracy only

    after training for few sessions. In this work

    90% of accuracy has been achieved, in

    classifying the two movements (RTL &

    LTR).

    ACKNOWLEDGEMENT

    The authors are indebted to the UGC. This work is a part of the funded major research project C.F. No 32-14/2006(SR)

    .

    REFERENCES

    1. The "10-20 System of Electrode Placement http://faculty.washington.edu/chudler/1020.html

    2. Y. U. Khan,(2010) Imagined wrist movement classification in single trial EEG for brain computer interface using wavelet

    packet, Int. J. Biomedical Engineering and Technology, Vol. 4, No. 2, pp169-180.

    3. Daniel, J. Szafir (2009-10) Non-Invasive BCI through EEG An Exploration of the Utilization of electroencephalography to Create

    Thought-Based Brain-Computer Interfaces.

    4. Wolpaw, J.R., Birbaumer, N., McFarland, D.J., Pfurtscheller, G., Vaughan, T.M. (2002): Braincomputer interfaces for communication and control. Clinical Neurophys. pp767791

    5. Y. U. Khan and O. Farooq(2009), Autoregressive features based classification for seizure detection using neural network in scalp

    Electroencephalogram, International Journal of Biomedical

    Engineering and Technology, vol.2, no. 4, pp. 370-381.

    6. J. Vidal(1973) "Toward Direct BrainComputer Communication." Annual Review of Biophysics and Bioengineering. Vol. 2, pp. 157-

    180

    7. Syed M.Siddique, Laraib Hassan Siddique (2009): EEG based Brain computer Interface: Journal of software, vol.4, no.6, pp.550-

    555

    8. EEG Channels in Detecting Wrist Movement Direction Intention: Proceedings of the 2004 IEEE Conference on Cybernetics and Intelligent Systems

    9. Fabiani, Georg E. et al. Conversion of EEG activity into cursor movement by a brain-computer interface.

    . 2004

    10. Clarity Braintech system, Standard edition, Software version 3.4, Hardware version 1.4, Clarity Medical Private Limited

    -15 -10 -5 0 5 10 15 2010

    20

    30

    40

    50

    60

    RTL

    LTR

    Support Vectors

    Classifier

    0 5 10 15 200

    200

    400

    600

    800

    1000

    Time(sec)

    Va

    ria

    nce

    Variance of FP2F4

    RTL

    LTR