Page 335 - Electrical Engineering Dictionary
P. 335

The scaling factor allows the same matrix to  These are known as the Sylvester matrices.
                              be used for the inverse transform. The 4 × 4  Because of their properties, Hadamard matri-
                              Haar transform can be interpreted as follows:  ces find application in the theory of error con-
                              first apply the 2 × 2 transform to two in-  trol codes, and code division multiple access.
                              dependent pairs of samples; then apply the  Named after Jacques Salomon Hadamard
                              2×2transformtothetwoaveragecoefficients  (1865–1963). See also orthogonal, Walsh
                              just computed. Larger Haar transforms are  cover.
                              constructed by continuing this process recur-
                              sively.                                Hadamard transform    a unitary trans-
                                The Haar transform yields coefficients  form mapping N samples g(n) to N coeffi-
                              equal to the subband values generated by  cientsG(k)accordingtothetransformmatrix
                              dyadic decomposition with the Haar wavelet.  H N , where
                              This transform has achieved rather less use           1    1  1
                              than the other transforms in this family,       H 2 = √   1 −1
                                                                                     2
                              such as the Discrete Cosine, Fourier and
                              Hadamard transforms.                   and larger arrays are formed by the recursive
                                                                     definition
                                                                                               !
                              Haar wavelet   the orthonormal wavelet              1   H N  H N
                                   1   1    1  −1                                       2    2
                              pair ( √ , √ ), ( √ , √ ). Analysis and syn-  H N = √              .
                                    2  2    2   2                                  2  H N −H N
                              thesis pairs are identical. This is the most              2    2
                              compact wavelet pair. Dyadic subband de-  The inverse transform is identical.  The
                              composition with the Haar wavelets yields  Hadamard transform was formerly used for
                              coefficients equal to those from the Haar  data compression because its entries are all
                              transform.                             1or −1, allowing computation without mul-
                                                                     tiplications. In this context it is now super-
                              Haas effect   states that the first sound  seded by the discrete cosine transform. See
                              heard will mask subsequent short delay ar-  also Walsh–Hadamard transform.
                              riving sounds, the combination appearing as
                              a louder source. Also called law of the first  half adder  a logic circuit thatproduces the
                              wavefront.                             sum and carry outputs for two input signals.
                                                                     A half adder has no carry input.
                              hacker   a person who explores computer
                                                                     half bridge amplifier  a class-D amplifier
                              and communication systems, usually for in-
                                                                     based on a half-bridge inverter configuration.
                              tellectual challenge, commonly applied to
                                                                     Not suitable for amplification of DC to low-
                              those who try to circumvent security barri-
                                                                     frequency signals because the capacitor leg
                              ers (crackers).
                                                                     cannot provide unidirectional current.
                              Hadamardmatrix    ann×nmatrixH with
                                                                     half subtracter  a logic circuit that pro-
                              elements ±1 is a Hadamard matrix of order n  vides the difference and borrow outputs for
                              if HH T  = nI, i.e., the rows are all mutually
                                                                     two input signals. A half subtracter has no
                              orthogonal, as are the columns. Hadamard
                                                                     borrow input.
                              matrices can only exist for n = 1, 2or n an
                              integer multiple of 4. Hadamard matrices of  half-band filter  a filter whose even-
                                   i
                              order 2 can be constructed by the recursion
                                                                     indexed coefficients are all zeros except the
                                                                     one in the filter center.
                                        H 1 = (1)

                                               H n H n               half-height point  a point at which the
                                       H 2n =
                                               H n −H n              membership grade is equal to 0.5.
                              c 
2000 by CRC Press LLC
   330   331   332   333   334   335   336   337   338   339   340