Multi User Detection

Embed Size (px)

Citation preview

  • 7/28/2019 Multi User Detection

    1/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    Multiuser Detection

    Mohit Garg

    ([email protected])

    Under the guidance ofProf. U. B. Desai

    SPANN Lab

    Department of Electrical EngineeringIIT-Bombay

    Group Members: Prof. S. N. Merchant

    Aditya Dua, Ritesh Sood, Prateek Dayal, Umesh Nimbhorkar

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    2/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    Outline ...

    Recapitulating CDMA

    Standard single user detector Optimum multiuser detector Non-optimal multiuser detectors

    An adaptive Minimum Probability of Error based multiuser detector pro-posed by our group

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    3/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    Code Division Multiple Access (CDMA)

    All users transmit at the same time and across the entire frequency band

    Users separated on the basis of their signature waveforms

    sk(t) The signature waveforms may be Orthogonal

    Non-orthogonal

    Orthogonal CDMA does not give any capacity improvement over TDMA

    or FDMA in a cellular system

    No. of orthogonal signature waveforms is limited!

    Non-orthogonal CDMA is therefore used

    http://prevpage/http://prevpage/http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/http://prevpage/
  • 7/28/2019 Multi User Detection

    4/75

  • 7/28/2019 Multi User Detection

    5/75

  • 7/28/2019 Multi User Detection

    6/75Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    The Single User Detector . . .

    The simplest way: Apply a bank of matched filters, one for each user, to the

    received signal

    Demodulate all users independent of each other Consider the first time interval (i = 0) and the jth user,

    r(t) = Ajbjsj(t) + Kk=1k=j

    Akbk(i)sk(t)

    + n(t), 0 t < Tb= Signal + M AI + Noise

    (6)

    Multiple Access Interference (MAI) In-band interference unlike noise which is wideband

    Cannot be rejected through a band-pass filter

    Occurs in different forms in other systems also e.g. Multi-Carrier in-

    terference in OFDM

    http://prevpage/http://prevpage/http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/http://prevpage/
  • 7/28/2019 Multi User Detection

    7/75

  • 7/28/2019 Multi User Detection

    8/75

  • 7/28/2019 Multi User Detection

    9/75Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    Issues with the Single User Detector

    + Simple to implement

    + Does not require knowledge of the channel or the user amplitudes

    Multiple Access Interference (MAI) Kk=1k=j

    Akbk(i)kj

    Gives non-zero probability of error even with zero noise due to MAI Near-Far Effect: Strong users overwhelm the weak ones. Thus stringent

    power control is necessary

    The single user detector forms the core of almost all CDMA handsets currently in use

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    10/75Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    Optimality of the Single User Detector

    The matched filter is optimum only for AWGN interference The MAI term is not Gaussian in general

    But can be approximated by a Gaussian random variable for large no.

    of users Central Limit Theorem

    Even if the MAI was Gaussian, the single user detector is still not optimum yj is not a sufficient statistic for bj

    But [y1, . . . , yK]T is a sufficient statistic for [b1, . . . , bK]T

    The single user detector would have been optimum Ifij = 0

    i, j

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    11/75Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    Therefore,

    Multiuser Detection!

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    12/75Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    Multiuser Detection

    It is clear that the single user detector is not optimum

    The optimum detector should take into account the information available inall yks to estimate the bit of a particular user This is known as Multiuser Detection and was proposed by Sergio Verdu in

    early 1980s.

    Any multiuser detector will utilise the information available in the MAI

    term to demodulate the user and will not treat it like a noise term

    Processing the interference term to extract useful information

    Ideologically similar to utilising multipaths for diversity!

    http://prevpage/http://prevpage/http://nextpage/http://nextpage/http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/http://nextpage/http://prevpage/
  • 7/28/2019 Multi User Detection

    13/75Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    Individually Optimum Multiuser Detector

    Consider the simple 2-user case

    r(t) = A1b1s1(t) + A2b2s2(t) + n(t), 0 t < Tb (10) The optimum estimate ofb1 will minimise the probability of error It is obtained by choosing b1 {1, +1} such that the aposteriori proba-

    bility P(b1

    |r(t), 0

    t < Tb) is maximised.

    Similarly for user 2, i.e. we need to choose b2 such that P(b2|r(t), 0 t

    ||y||2y

    http://prevpage/http://prevpage/http://nextpage/http://nextpage/http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/http://nextpage/http://prevpage/
  • 7/28/2019 Multi User Detection

    48/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    Reducing Complexity Further . . .To further reduce the computational cost we consider:

    MPOE Implementation using ISI Cancellation: Efficient MJPOE Minimizing the probability of error for each user separatelyMCPOE

    http://prevpage/http://prevpage/http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/http://prevpage/
  • 7/28/2019 Multi User Detection

    49/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    Efficient MJPOEA proposed modification to the MJPOE algorithm proposed will now be dis-

    cussed which

    Reduces the computation allowing pre-computation of weights Weight computation is now done on channel variation timescales rather than

    symbol timescales

    Improves the BER performance Using the same signal model as before, the received signal at thepth antenna

    can be written as

    rp(i) = SpLAb(i 1) + SpRAb(i) + np

    Demodulation

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    50/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    Demodulation

    The receiver structure consists of a linear filter.

    Augmented received vector r = [rT1 , . . . , rTP]

    T (N P 1) Weights for the kth user, wk = [w

    Tk1, . . . ,w

    TkP]

    T (N P 1) . Filter Output at the ith bit interval is

    yk(i) = wH

    kr(i)

    =

    Pp=1

    wHkprp

    =P

    p=1 wHkpSpLAb(i 1) +P

    p=1 wHkpSpRAb(i) + wHk n The Decision Rule is

    bk = sgn[

    (yk)]

    Modification

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    51/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    Modification

    E(

    [yk]) =

    [k]

    = 0

    Hence, a threshold of0 is not optimum Subtracting an estimate of this term from yk, we obtain the proposed deci-

    sion statistic zk

    zk = yk kwhere,

    k =

    P

    p=1 wHkpSpLAb(i 1) Since, E([zk]) = 0, we get the proposed decision rule as

    bk = sgn[(zk)]

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    52/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    Modification . . .

    Some observations on the proposed demodulation scheme Does not involve any ISI term. Thus, we can pre-compute the weights

    for MJPOE

    (Cannot pre-compute for MCPOE since it requires training)

    E([zk] ) = 0, which is the decision threshold. Thus, reduction inBER is expected

    Simulation results support the claim

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    53/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    MCPOE: Conditional Probability of Error

    http://prevpage/http://prevpage/http://nextpage/http://nextpage/http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/http://nextpage/http://prevpage/
  • 7/28/2019 Multi User Detection

    54/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    MCPOE: Conditional Probability of Error

    If bits +1 and -1 are equiprobable, the probability of error for the

    k

    th user,

    conditioned on the transmitted bit vector b given by :

    Pk|b =1

    2P((yk) < 0|bk = 1) + 1

    2P((yk) 0|bk = 1)

    Density function ofyk conditioned on b given by:f(yk)|b(y|b) =

    1

    k

    2exp

    (y k)

    2

    22k

    where:

    k = k + Pp=1

    wHkpSpRAb

    k = wHk wk

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    55/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    Conditional Probability of Error . . .

    Let k = k|1 when bk = 1 and k = k|1 when bk = 1

    k|1 = k + Pp=1

    wHkp Ki=1

    i=k

    Aibi

    Mm=1

    aim,pgimsiL(nim) + Ak

    Mm=1

    akm,pgkmskL(nkm)

    k|1 =

    k +

    P

    p=1wHkp

    K

    i=1i=k AibiM

    m=1aim,pgimsiL(nim) Ak

    M

    m=1 akm,pgkmskL(nkm)

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    56/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    The conditional probability of error can be expressed as :

    Pk|b =1

    2

    0

    1

    k

    2exp

    (y k|1)

    2

    22k

    dy

    +1

    2

    0

    1

    k2exp

    (y k|1)2

    22k dy The above expression can be simplified to obtain :

    Pk|b =1

    2+

    1

    2Q

    k|1k

    12

    Q

    k|1

    k Minimize Pk|b with respect to wk

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    57/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    Figure 3: Schematic of the Space-Time MCPOE Adaptive Detector

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    58/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    MCPOE Adaptive Algorithm MCPOE : Minimum Conditional Probability of Error Minimizes probability of error for each user (conditioned on

    transmitted bit vector) individually

    Training sequence required for adaptation Gradient descent based adaptation of filter weights Ifw(i)k denotes the filter for detecting the kth user during the ith bit interval,

    then :

    w(i+1)k = w

    (i)k Pk|bwk wk=w(i)k

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    59/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    MCPOE Adaptive Algorithm . . .

    Derivative of

    Pk|bw.r.t w

    kcan be computed using:

    Pk|bwk

    =12

    exp

    2k|1

    2k

    wk

    k|1k

    1

    2exp2k|1

    2k

    wkk|1

    k

    MCPOE Adaptive Algorithm

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    60/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    MCPOE Adaptive Algorithm . . .

    Define ukp = k|1wkp

    and vkp = k|1wkp

    uk = [uHk1, . . . ,uHkP]H and vk = [vHk1, . . . ,vHkP]H

    ukp = SpLAb(i 1) +K

    j=1j=k

    Ajbj

    M

    m=1 ajm,pgjms(njm)jR + AkM

    m=1akm,pgkms(nkm)kR vkp =

    SpLAb(i 1) + Kj=1

    j=k

    Ajbj

    Mm=1

    ajm,pgjms(njm)jR Ak

    Mm=1

    akm,pgkms(nkm)kR

    wk

    k|1k

    = wk2uk k|1wkwk3/2

    wk

    k|1

    k

    =

    wk2vk k|1wkwk3/2

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    61/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    MCPOE Adaptive Algorithm . . .Blind Version

    MCPOE requires a training sequence in adaptation mode We tried to make it blind by conditioning the probability of error

    expression on the output of a matched filter rather than on the

    training bits

    Convergence rate not affected by the modification

    Slight degradation in BER performance Similar modification works poorly for other training based adaptivedetectors such as LMS and RLS

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    62/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    Figure 4: Schematic of the Blind MCPOE Adaptive Detector : Output of Matched filter Bank used as training bits

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    63/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    Simulation Results

    Results for both synchronous and asynchronous multipath channels (with

    multiple antennas)

    BER performance compared with non-adaptive MMSE Convergence rates compared with training based LMS and RLS

    Flat Rayleigh faded channel assumed for simulations

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    64/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen QuitFigure 5: Convergence curves for training based LMS and RLS (Multipath)

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    65/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen QuitFigure 6: Convergence Curves for MJPOE (Multipath)

    100

    Convergence Curves at 15dB Transmit SNR

    Theoretical Probability of Error: MJPOE

    http://prevpage/http://prevpage/http://nextpage/http://nextpage/http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/http://nextpage/http://prevpage/
  • 7/28/2019 Multi User Detection

    66/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    0 50 100 15010

    3

    102

    101

    No. of Iterations

    Prob.ofEr

    ror(on

    logscale)

    Theoretical Probability of Error: MJPOETheoretical Probability of Error: Efficient MJOPE

    Figure 7: Convergence performance comparison of MJPOE and Efficient MJPOE, K = 4, M = 3, N = 15, P = 4

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    67/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    Figure 8: Schematic of the Diversity Combining MCPOE Adaptive Detector

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    68/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    Figure 9: Convergence Curves for Training Based MCPOE (Multipath)

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    69/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    Figure 10: Comparison of MCPOE and Non-Adaptive MMSE (Multipath)

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    70/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    Figure 11: BER performance comparison of MJPOE, MCPOE and MMSE

    100 Prob. of Error Vs. SNR

    Simulated Recursive Least Squares (RLS)

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    71/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    0 2 4 6 8 10 1210

    4

    103

    102

    101

    Transmit SignaltoNoise Ratio (SNR) (in dB)

    Prob.ofEr

    ror(on

    logscale)

    Simulated Recursive Least Squares (RLS)Theoretical Prob. of Error: MJPOETheoretical Prob. of Error: Efficient MJPOESimulated Prob. of Error: Efficient MJPOE

    Figure 12: BER performance comparison of MJPOE and Efficient MJPOE, K = 4, M = 3, N = 15, P = 4

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    72/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    Figure 13: Effect of number of antennas on performance of MJPOE, for a fixed number of multipaths

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    73/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    Conclusion

    + MPOE based adaptive multiuser detection algorithms were proposed

    + Performance better than MMSE based approaches

    + Similar to the optimum multiuser detector without the overhead of expo-nential computation at each bit interval

    High computational complexity

    Work has been extended to OFDM-SDMA and MC-CDMA with encourag-ing results

    References

    http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/
  • 7/28/2019 Multi User Detection

    74/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    Sergio Verdu, Multiuser Detection, Cambridge University Press, 1998.

    Aditya Dua, U.B. Desai and R.K. Mallik, Minimum probability of error-based methods for adaptive multiuser detection in multipath DS-CDMA

    channels, IEEE Transactions on Wireless Communications, May 2004.

    R. Sood and U. B. Desai, Minimum probability of error demodulation for

    multipath OFDM-SDMA systems, IEEE International Conference on Com-

    munications, Jun. 2004.

    P. M. Dayal, U. B. Desai and A. Mahanta, Minimum conditional probabil-ity of error detection for MC-CDMA, IEEE International Symposium on

    Spread Spectrum Techniques and Applications, Aug. 30-Sept. 2 2004.

    Mohit Garg, Umesh Nimbhorkar, U. B. Desai and S. N. Merchant, Efficientminimum probability of error demodulation for DS-CDMA systems, To ap-

    pear in IEEE Wireless Communications and Networking Conference, Mar.

    2005.

    Prof. U. B. Desai

    [email protected]

    SPANN Lab

    Dept. of Electrical Engineering

    http://prevpage/http://prevpage/http://nextpage/http://nextpage/http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/http://nextpage/http://prevpage/
  • 7/28/2019 Multi User Detection

    75/75

    Dec. 06, 2004 Mohit Garg ISTE-STTP, TSEC First Prev Next Last Go ToFull Screen Quit

    p g g

    Indian Institute of Technology - Bombay

    Mumbai 400076

    Mohit Garg

    [email protected]

    Dept. of Electrical Engineering

    Indian Institute of Technology - Bombay

    Mumbai 400076

    http://prevpage/http://prevpage/http://nextpage/http://nextpage/http://lastpage/http://lastpage/http://fullscreen/http://fullscreen/http://quit/http://quit/http://quit/http://fullscreen/http://lastpage/http://nextpage/http://prevpage/