Acoustic Beamforming for Signal Enhancement, Localization, and by wio18411

VIEWS: 21 PAGES: 7

									    Acoustic Beamforming for Signal
Enhancement, Localization, and Separation

              Prof. Kung Yao
                  UCLA

   DARPA Air-Coupled Acoustic Sensors Workshop
              August 24-25, 1999
                Crystal City, VA
    Advances in Beamforming Concepts

Beamforming based on coherent processing of
acoustic sensor data can be used to:
 1. Enhance acoustic signal in known / unknown
    direction/position
 2. Localize the acoustic source(s) in 2-D or 3-D
 3. Blindly separate different acoustic sources due
    to their spatial separations and not their
    frequency characteristics
        Practical Implementation of
     Acoustic Beamforming Techniques

We have conducted research in these directions:
1. Steerable microphone array to enhance speaker/
   reject interferers(s) for hearing aid appl. (HEI)
2. Blind beamforming of randomly placed
   sensors for enhanced detection/classification;
   localization/tracking of vehicles/foot steps
   (DARPA/ATO AWAIRS project with RSC)
3. Blind separation of two (or more) speakers
   using subspace blind deconvolution/decorrelation
 Microphone Array for Hearing Aid App.
• Use four microphones in a linearly array
  with about 2.2 cm inter-sensor separation
• Use “maximum power”criterion beamformer
  to steer mainlobe toward desired speaker and
  nulls toward interferers
Case   Source/Angle   Interfer/Angle   SIR(dB)   ∆SNR(db)Meas.
  1    Speech/0       Speech/-45        0         0 No processing
  2    Speech/0       Speech/-45        0        22 Free space
  3    Speech/0       Speech/-45        0        12 Low-rev.
  4    Speech/0       C.noise/-45      -10        0 No processing
  5    Speech/0       C.noise/-45      -10       21 Free space
  6    Speech/0       C.noise/-45      -10       12 Low-rev.
 Estimation of a Heavy Track Vehicle Using
An AWAIRS Blind Beamforming Algorithm
  E s t im a t e d H e a vy V e h ic le Tra c k R e c o rd (u s in g t im e -d e la y s fro m s ix s e n s o rs in R U N 3 )
  5200



  5150



  5100

                                                      4
                                                           11
  5050                                                          7
                                                  2        3
                                                      14

  5000



  4950



  4900



  4850
     4900               4950              5000                 5050          5100              5150              5200
                       Blind Source Separator System
                                                  L-tap FIR
                                                  deconvolver
 d1(n)
                    h1,1(n)
                                                  e1,1 (p)(n)
                                         x1(n)
                                   A/D                                y1(n)                   s1(n)
                        h1,2(n)
                                                  e2,1   (p)(n)
                                                                  +
                      h2,1(n)




                                                                               decorrelator
                                                  e1,2 (p)(n)
                                         x2(n)
d2(n)                              A/D
                     h2,2(n)
                                                  e2,2 (p)(n)

                                                                       y2(n)                  s2(n)

                                    :                             +
 :
 :                      hNs,1(n)
                                    :
         hNs,2(n)
dNt(n)                                            e1,Ns (p)(n)
                                         xNs(n)
                                   A/D
                                                  e2,Ns (p)(n)

								
To top