Quadrature-Mirror Filter Bank

Transcription

1 Quadrature-Mirror Filter Bank In many applications, a discrete-time signal x[n] is split into a number of subband signals { v k [ n]} by means of an analysis filter bank The subband signals are then processed Finally, the processed subband signals are combined by a synthesis filter bank resulting in an output signal y[n]

2 Quadrature-Mirror Filter Bank If the subband signals { v k [ n]} are bandlimited to frequency ranges much smaller than that of the original input signal x[n], they can bedown-sampled before processing Because of the lower sampling rate, the processing of the down-sampled signals can be carried out more efficiently

3 Quadrature-Mirror Filter Bank After processing, these signals are then upsampled before being combined by the synthesis filter bank into a higher-rate signal The combined structure is called a quadrature-mirror filter (QMF) bank 3

4 Quadrature-Mirror Filter Bank If the down-sampling and up-sampling factors are equal to or greater than the number of bands of the filter bank, then the output y[n] can be made to retain some or all of the characteristics of the input signal x[n] by choosing appropriately the filters in the structure 4

5 Quadrature-Mirror Filter Bank If the up-sampling and down-sampling factors are equal to the number of bands, then the structure is called a critically sampled filter bank The most common application of this scheme is in the efficient coding of a signal x[n] 5

6 Two-Channel QMF Bank Figure below shows the basic two-channel QMF bank-based subband codec (coder/decoder) 6

7 Two-Channel QMF Bank The analysis filters H 0 (z) and H (z) have typically a lowpass and highpass frequency responses, respectively, with a cutoff at π/ 7

8 Two-Channel QMF Bank Each down-sampled subband signal is encoded by exploiting the special spectral properties of the signal, such as energy levels and perceptual importance It follows from the figure that the sampling rates of the output y[n] and the input x[n] are the same 8

9 Two-Channel QMF Bank 9 The analysis and the synthesis filters are chosen so as to ensure that the reconstructed output y[n] is a reasonably close replica of the input x[n] Moreover, they are also designed to provide good frequency selectivity in order to ensure that the sum of the power of the subband signals is reasonably close to the input signal power

10 Two-Channel QMF Bank 0 In practice, various errors are generated in this scheme In addition to the coding error and errors caused by transmission of the coded signals through the channel, the QMF bank itself introduces several errors due to the sampling rate alterations and imperfect filters We ignore the coding and the channel errors

11 Two-Channel QMF Bank We investigate only the errors caused by the sampling rate alterations and their effects on the performance of the system To this end, we consider the QMF bank structure without the coders and the decoders as shown below

12 Analysis of the Two-Channel QMF Bank Making use of the input-output relations of the down-sampler and the up-sampler in the z-domain we arrive at V ( z) H ( z) X ( z), U k = k / / k ( z) = { V ( z ) + V ( z )}, k = 0, k ^ k ( z) = V U k k ( z )

13 Analysis of the Two-Channel QMF Bank From the first and the last equations we obtain after some algebra ^ k V ( z) { V ( z) + V ( z)} = = k { H ( z) X ( z) + H ( z) X ( z)} k k k 3 The reconstructed output of the filter bank is given by ^ ^ ( z) = G ( z) V ( z) G ( z) V ( z) Y 0 0 +

14 Analysis of the Two-Channel QMF Bank From the two equations of the previous slide we arrive at 4 Y H0( z) G0( z) + H( z) G ( z) = { H ( z) G ( z) + H ( z) G ( z)} X ( z) { ( z)} X ( z) The second term in the above equation is due to the aliasing caused by sampling rate alteration

15 Analysis of the Two-Channel QMF Bank 5 The input-output equation of the filter bank can be compactly written as Y( z) = T( z) X ( z) + A( z) X ( z) where T(z), called the distortion transfer function, is given by T 0 0 ( z) = { H ( z) G ( z) + H ( z) G ( z)} and ( z) = { H ( z) G ( z) + H ( z) G ( z)} A 0 0

16 Alias-Free Filter Bank 6 Since the up-sampler and the down-sampler are linear time-varying components, in general, the -channel QMF structure is a linear time-varying system It can be shown that the -channel QMF structure has a period of However, it is possible to choose the analysis and synthesis filters such that the aliasing effect is canceled resulting in a time-invariant operation

17 Alias-Free Filter Bank 7 To cancel aliasing we need to ensure that A(z) = 0, i.e., H ( z) G ( z) + H ( z) G ( z) = For aliasing cancellation we can choose G ( z) H ( z) 0 = G ( z) H0( z) This yields G ( z) C( z) H ( z), G ( z) = C( z) H ( z), 0 = 0 where C(z) is an arbitrary rational function

18 Alias-Free Filter Bank 8 If the above relations hold, then the QMF system is time-invariant with an inputoutput relation given by Y(z) = T(z)X(z) where T( z) = { H ( z) H ( z) + H ( z) H ( z)} 0 0 On the unit circle, we have jω jω jω jω jφ ω) Y ( e ) = T( e ) X ( e ) = T( e ) e X ( e ( jω )

19 9 Alias-Free Filter Bank If T(z) is an allpass function, i.e., T( e j ) = with d 0 then jω jω Y ( e ) = d X ( e ) indicating that the output of the QMF bank has the same magnitude response as that of the input (scaled by d) but exhibits phase distortion The filter bank is said to be magnitude preserving ω d

20 Alias-Free Filter Bank If T(z) has linear phase, i.e., jω arg{ T( e )} = φ( ω) = αω + β then jω jω arg{ Y( e )} = arg{ X ( e )}+ αω + β The filter bank is said to be phasepreserving but exhibits magnitude distortion 0

21 Alias-Free Filter Bank If an alias-free filter bank has no magnitude and phase distortion, then it is called a perfect reconstruction (PR) QMF bank l In such a case, T ( z) = dz resulting in l Y( z) = dz X ( z) In the time-domain, the input-output relation for all possible inputs is given by y[ n] = d x[ n l]

22 Alias-Free Filter Bank Thus, for a perfect reconstruction QMF bank, the output is a scaled, delayed replica of the input Example - Consider the system shown below x[n] z z + y[n]

23 Alias-Free Filter Bank 3 Comparing this structure with the general QMF bank structure we conclude that here we have H ( z) =, H ( z) = z, G ( z) = z, G ( z) 0 0 = Substituting these values in the expressions for T(z) and A(z) we get ( z + z = ( z ) = 0 T ( z) = ) z A( z) = z

24 Alias-Free Filter Bank Thus the simple multirate structure is an alias-free perfect reconstruction filter bank However, the filters in the bank do not provide any frequency selectivity 4

25 An Alias-Free Realization 5 A very simple alias-free -channel QMF bank is obtained when H ( z) = H 0 ( z) The above condition, in the case of a real coefficient filter, implies jω j( π ω) H( e ) = H0( e ) indicating that if H 0 (z) is a lowpass filter, then H (z) is a highpass filter, and vice versa

26 An Alias-Free Realization jω j( π ω) The relation H( e ) = H0( e ) jω indicates that H ( e ) is a mirror-image jω of H 0 ( e ) with respect to π/, the quadrature frequency This has given rise to the name quadraturemirror filter bank 6

27 An Alias-Free Realization 7 Substituting H ( z) = H 0 ( z) in G0( z) = C( z) H( z), G( z) = C( z) H0( z), with C(z) = we get G0( z) = H( z), G( z) = H( z) = H0( z) The above equations imply that the two analysis filters and the two synthesis filters are essentially determined from one transfer function H 0 (z)

28 An Alias-Free Realization Moreover, if H 0 (z) is a lowpass filter, then G 0 (z) is also a lowpass filter and G (z) is a highpass filter The distortion function in this case reduces to T( z) = { H ( z) H ( z)} = { H ( z) H ( z)}

29 An Alias-Free Realization A computationally efficient realization of the above QMF bank is obtained by realizing the analysis and synthesis filters in polyphase form Let the -band Type polyphase representation of H 0 (z) be given by H ( z) = E ( z ) + z E ( ) 0 0 z 9

30 An Alias-Free Realization Then from the relation H ( z) = H 0 ( z) it follows that H( z) = E0( z ) z E( z ) Combining the last two equations in a matrix form we get H H 0 0 ( z) ( z) = 0( ) E z z E ( z ) 30

31 An Alias-Free Realization Likewise, the synthesis filters can be expressed in a matrix form as [ ] = [ G ] Making use of the last two equations we can 0( z) G ( z) z E( z ) E0( z ) redraw the two-channel QMF bank as shown below 3

32 An Alias-Free Realization Making use of the cascade equivalences, the above structure can be further simplified as shown below 3

33 An Alias-Free Realization Substituting the polyphase representations of the analysis filters we arrive at the expression for the distortion function T(z) in terms of the polyphase components as T( z) = z E ( z ) E ( z 0 ) 33

34 An Alias-Free Realization 34 Example - Let Its polyphase components are Hence H ) = + z 0( z H( z) = H0( z) = E0( z ) z E( z ) = z Likewise G ( z) = z E ( z ) + E ( z ) = + z E0 ( z ) =, E( z ) = 0 0 ( z) z E( z ) E0( z = G = ) + z

35 An Alias-Free Realization The distortion transfer function for this realization is thus given by T ( z) = z E ( z ) E ( z ) = z 0 The resulting structure is a perfect reconstruction QMF bank 35

36 Alias-Free FIR QMF Bank 36 If in the above alias-free QMF bank H 0( z ) is a linear-phase FIR filter, then its polyphase components E 0( z ) and E ( z), are also linear-phase FIR transfer functions In this case, T ( z) = z E ( ) ( 0 z E z ) exhibits a linear-phase characteristic As a result, the corresponding -channel QMF bank has no phase distortion

37 Alias-Free FIR QMF Bank 37 However, in general T ( ) is not a constant, and as a result, the QMF bank exhibits magnitude distortion We next outline a method to minimize the residual amplitude distortion Let H 0( z ) be a length-n real-coefficient linear-phase FIR transfer function: H e j ω N = 0 ( z) n= 0h0[ n] z n

38 Alias-Free FIR QMF Bank 38 Note: H 0( z ) can either be a Type or a Type linear-phase FIR transfer function since it has to be a lowpass filter Then h 0[ n ] satisfy the condition h0[ n] = h0[ N n] In this case we can write H ( ) / 0 e j ω = e j ω N ~ H0( ω) In the above H~ 0( ω ) is the amplitude function, a real function of ω

39 Alias-Free FIR QMF Bank 39 The frequency response of the distortion transfer function can now be written as jnω T ( e jω) = e { H ( ω) ( ) ( ( π ω) ) 0 e j N H j 0 e } From the above, it can be seen that if N is even, then T ( e jω ) = 0 at ω = π/, implying severe amplitude distortion at the output of the filter bank

40 Alias-Free FIR QMF Bank 40 N must be odd, in which case we have jnω T e jω ) = e { H ( ) ( ( ) 0 e jω + H0 e j π ω ) ω = e jn { H ( ) ( ) 0 e jω + H e jω } It follows from the above that the FIR - channel QMF bank will be of perfect reconstruction type if H ( e jω) + H ( e jω) ( 0 = }

41 Alias-Free FIR QMF Bank 4 Now, the -channel QMF bank with linearphase filters has no phase distortion, but will always exhibit amplitude distortion jω unless T( e ) is a constant for all ω If H 0 ( z) is a very good lowpass filter with H0( e jω ) in the passband and H0( e jω ) 0 in the stopband, then H ( z ) is a very good highpass filter with its passband coinciding with the stopband of H 0( z ), and vice-versa

42 Alias-Free FIR QMF Bank jω As a result, T( e ) / in the passbands of H 0( z ) and H ( z) Amplitude distortion occurs primarily in the transition band of these filters Degree of distortion determined by the amount of overlap between their squaredmagnitude responses 4

43 Alias-Free FIR QMF Bank 43 This distortion can be minimized by controlling the overlap, which in turn can be controlled by appropriately choosing the passband edge of H 0( z ) One way to minimize the amplitude distortion is to iteratively adjust the filter coefficients h 0 [n] of H 0 ( z) on a computer such that jω jω H0( e ) + H( e ) is satisfied for all values of ω

44 Alias-Free FIR QMF Bank To this end, the objective function φ to be minimized can be chosen as a linear combination of two functions: () stopband attenuation of H 0 ( z), and () sum of squared magnitude responses of H ( ) and H (z) 0 z 44

45 Alias-Free FIR QMF Bank 45 One such objective function is given by φ = αφ+ ( α) φ where π φ H e jω = ( ) d ω ω and s π jω jω φ = H0( e ) H( e ) dω 0 and 0 < α <, and ωs = π + ε for some small ε > 0

46 Alias-Free FIR QMF Bank 46 Since T ( ) is symmetric with respect to π/, the second integral in the objective function φ can be replaced with φ e j ω π/ ω ω = H0( e j ) H( e j ) 0 dω After φ has been made very small by the optimization procedure, both φ and φ will also be very small

47 Alias-Free FIR QMF Bank Using this approach, Johnston has designed a large class of linear-phase FIR filters meeting a variety of specifications and has tabulated their impulse response coefficients Program 0_9 can be used to verify the performance of Johnston s filters 47

48 Alias-Free FIR QMF Bank Example- The gain responses of the length- linear-phase FIR lowpass filter B and its power-complementary highpass filter obtained using Program 0_9 are shown below 0 0 H 0 (z) H (z) Gain, db ω/π

49 Alias-Free FIR QMF Bank The program then computes the amplitude distortion H 0 ( e j ω ) + H( e j ω) in db as shown below Amplitude distortion in db ω/π

50 Alias-Free FIR QMF Bank From the gain response plot it can be seen that the stopband edge ω s of the lowpass filterb is about 0.7π, which corresponds to a transition bandwidth of ω 0.5π) / = 0. 05π ( s The minimum stopband attenuation is approximately 34 db 50

51 Alias-Free FIR QMF Bank The amplitude distortion function is very close to 0 db in both the passbands and the stopbands of the two filters, with a peak value of ± 0.0 db 5

52 Alias-Free IIR QMF Bank 5 Under the alias-free conditions of G0( z) = H( z), G ( z) = H0( z) and the relation H( z) = H0( z), the distortion function T(z) is given by T ( z) = z E ( ) ( 0 z E z ) If T(z) is an allpass function, then its magnitude response is a constant, and as a result its corresponding QMF bank has no magnitude distortion

53 Alias-Free IIR QMF Bank 53 Let the polyphase components E 0 ( z) and E ( z ) of H 0( z ) be expressed as E z) = A ( ), E z) = A ( ) 0( 0 z ( z with A 0( z) and A ( z) being stable allpass functions Thus, H ( z) = H [ A ( z) + z A ( )] 0 0 z ( ) ( ) ( z = A 0 z z A z [ )]

54 Alias-Free IIR QMF Bank 54 In matrix form, the analysis filters can be expressed as H ( ) ( 0 z = A0 z ) H ( ) ( 0 z z A z ) The corresponding synthesis filters in matrix form are given by G ( ) ( ) = ( ) ( ) 0 z G z z A z A0 z [ ] [ ]

55 Alias-Free IIR QMF Bank Thus, the synthesis filters are given by ( ) ( ) ( z = [ A0 z + z A z )] = H( z G 0( z) = [ A0( z ) + z A( z )] = H0( z) G The realization of the magnitude-preserving -channel QMF bank is shown below ) 55

56 Alias-Free IIR QMF Bank From H ( ) [ ( ) ( ] 0 z = A0 z + z A z ) it can be seen that the lowpass transfer function H 0 (z) has a polyphase-like decomposition, except here the polyphase components are stable allpass transfer functions 56

57 Alias-Free IIR QMF Bank 57 It has been shown earlier that a boundedreal (BR) transfer function H0 ( z) = P0 ( z)/ D( z) of odd order, with no common factors between its numerator and denominator, can be expressed in the form H ( ) [ ( ) ( ] 0 z = A0 z + z A z ) if it satisfies the power-symmetry condition H0 ( z) H0( z ) + H0( z) H0( z ) = and has a symmetric numerator P 0 (z)

58 Alias-Free IIR QMF Bank 58 It has also been shown that any odd-order elliptic lowpass half-band filter H 0 (z) with a frequency response specification given by jω δ p H( e ), for 0 ω ω p j H( e ω ) δ s, for ωs ω π and satisfying the conditions ω p + ω s = π and δ s = 4δ p ( δ p ) can always be expressed in the form H [ A ( z) + z A ( )] 0 0 z ( z) =

59 Alias-Free IIR QMF Bank The poles of the elliptic filter satisfying the two conditions on bandedges and ripples lie on the imaginary axis Using the pole-interlacing property discussed earlier, on can readily identify the expressions for the two allpass transfer functions A 0 (z) and A (z) 59

60 Alias-Free IIR QMF Bank Example - The frequency response specifications of a real-coefficient lowpass half-band filter are given by: ω p= 0. 4π, ωs= 0. 6π, and δ s = From δ s = 4δ p ( δ p ) we get δ p = In db, the passband and stopband ripples are Rp = and Rs =

61 Alias-Free IIR QMF Bank Using the M-file ellipord we determine the minimum order of the elliptic lowpass filter to be 5 Next, using the M-file ellip the transfer function of the lowpass filter is determined whose gain response is shown below 0 Real half-band filter Gain, db ω/π

62 Alias-Free IIR QMF Bank The poles obtained using the function tfzp are at z = 0, z = ± j , and z = ± j The pole-zero plot obtained using zplane is shown below Imaginary Part Real Part

63 Alias-Free IIR QMF Bank Using the pole-interlacing property we arrive at the transfer functions of the two allpass filters as given below: z A 0( z ) = z A ( z z ) = z 63

64 64