• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.44 no.4 s.316
• /
• pp.70-77
• /
• 2007

Two kinds of QMF(Quadrature Mirror Filter) pairs are used in JPEG2000 standard, which don't have QMF distortions. However, the QMF pairs have the main disadvantages such that there are gentle roll-off rate, ripples in the passband and unequal band decomposition. In this paper, Maxflat(maximally flat) QMF pairs with a half-band gain are proposed for overcoming these problems. Maxflat QMF pairs are realized due to generalized closed-form formulas, and the filters have maximally flat response in the passband/stopband as well as sharp roll-off rate in the transition band. Comparing proposed filters and JPEG2000's filters in frequency domain, it is found that proposed filters have better performance JPEG2000's filters. Moreover, Maxflat QMF pairs show stopband-attenuation exceeding 200 dB almost everywhere.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.29 no.2C
• /
• pp.213-221
• /
• 2004

A new method for design of two-channel finite-impulse response(FIR) quadrature mirror-image filter(QMF) banks with low reconstruction delay using weighting function is proposed. The weighting function used in this paper is calculated from the previous updated filter coefficients vector which is adjusted from iteration to iteration in the design of QMF banks. In this paper, passband and stopband edge frequency are used in design of QMF banks with low delay characteristic in time domain instead of specific frequency interval where the artifacts occur in conventional design method. The investigation of specific frequency interval where artifacts occur can not be required by using passband and stopband edge frequency. Some comparisons of performance are made with other existing design method to demonstrate the proposed method for QMF bank design. and it was observed that the proposed method using the weighted function and passband and stopband edge frequency improves the peak reconstruction error by 0.001 [dB], the peak-to-peak passband ripple by 0.003[dB], SNR with a white noise by 7[dB] and SNR with a step input by 32[dB], but with a reduction of the computational efficiency because of updating the weighting function over the conventional method in Ref [11].

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.33 no.3C
• /
• pp.271-280
• /
• 2008

M channel near-perfect-reconstruction(NPR) pseudo-QMF banks are a hybrid of conventional pseudo-QMF design and spectral factorization approach where the analysis and synthesis filters are cosine-modulated versions of the prototype-lowpass filter(p-LPF). However, p-LPF H(z) does not have linear-phase symmetry as well as magnitude-distortion optimization since it is obtained by spectral factorization of $2M^{-th}$ band filter $G(z)=z^{-(N-1)}H(z^{-1})H(z)$. A fair amount of attention, therefore, has been focused on the design of filter banks for reducing only alias-cancellation distortion without reconstructed-amplitude distortion. In this paper, we propose a new method for designing linear-phase p-LPF in NPR pseudo-QMF banks, which is based on Maxflat(maximally flat) FIR filters with closed-form transfer function. In addition, p-LPF H(z) is optimized in this approach so that the 2M-channel overall distortion response represented with $G(z)=H^2(z)$ approximately becomes an unit magnitude response. Through several examples of NPR pseudo-QMF banks, it is shown that the peek ripple of the overall magnitude distortion is less than $3.5{\times}10^{-4}\;({\simeq}-70dB)$ and analysis/synthesis filters have the sharp monotone-stopband attenuation exceeding 100 dB.

• Journal of Broadcast Engineering
• /
• v.4 no.2
• /
• pp.144-154
• /
• 1999

The binomial sequences are family of orthogonal sequences that can be generated with remarkable simplicity-no multiplications are necessary. This paper introduces a class of non-recursive multidimensional filters for frequency-selective image processing without multiplication operations. The magnitude responses are narrow-band. approximately gaussian-shaped with center frequencies which can be positioned to yield low-pass. band-pass. or high-pass filtering. Algorithms for the efficient implementation of these filters in software or in hardware are described. Also. we show that the binomial QMFs are the maximally flat magnitude square Perfect Reconstruction paraunitary filters with good compression capability and these are shown to be wavelet filters as well. In wavelet transform the original image is decomposed at different scales using a pyramidal algorithm architecture. The decomposition is along the vertical and horizontal direction and maintains constant the number of pixels required to describe the images. An efficient perfect reconstruction binomial QMF-Wavelet signal decomposition structure is proposed. The technique provides a set of filter solutions with very good amplitude responses and band split. The proposed binomial QMF-filter structure is efficient, simple to implement on VLSl. and suitable for multi-resolution signal decomposition and coding applications.

• Proceedings of the Korea Institute of Convergence Signal Processing
• /
• 2000.12a
• /
• pp.85-88
• /
• 2000

Signal splitting and perfect reconstruction in subband coding is based on the assumption that quantization errors are negligible. But if subband signal is quantized, 4 types of errors occurs thus it is not impossible to do perfect reconstruction. These errors are QMF design error, aliasing error, signal error and random error. By using the QMF for subband splitting, the QMF error does not present. and by using the Lloyd-Max quantizer for the quantization and by using an appropriate synthesis filter, all signal dependent errors can be cancelled and the remaining error is random error which is uncorrelated with the original image〔1〕. In this thesis, Lenna and Camera-Man image are devided into 10 subbands by using the D4 and D20 wavelet And the subband signals are quantized by using the Lloyd-Max quantizer and the quantization errors are compared. and evaluated.

• Proceedings of the KIEE Conference
• /
• 1999.07g
• /
• pp.3221-3223
• /
• 1999

OAEs are low-level sounds produced by cochlea as part of the normal hearing process. OAEs can be measured with a microphone placed in the outer ear-cannel. We can diagnose cochlea's condition by using OAEs. To diagonose it's condition, however, is difficult by reason of OAEs' tiny. Thus, It need to a method using latency which essential diagnosing a time-element of OAEs. This study proposes a latency detection algorithm using 7-QMF for more effective detection of latency, 7-QMF designed by wavelet theory can process signal without a losses of information. The latency-detector based on 7-QMF is superior to former method.

• Journal of Korea Multimedia Society
• /
• v.2 no.1
• /
• pp.30-37
• /
• 1999

A wideband active noise cancelling system involves adaptive filters with hundreds of taps. The computational burden required with these long adaptive filters. This paper presents active noise cancelling system using M-channel QMF filter banks in which the adaptive weights are computed in subbands. The analysis and synthesis filter banks use cosine-modulated pseudo QMF filters. The reference signal for on-line identification of error path transfer characteristics is used to difference signal between the output of adaptive filters and the output of lowpass subband filters. The proposed adaptive subband filter bank suggests robust active noise cancelling system retaining the computational complexity and convergence speed advantaged of subband processing.

• Proceedings of the Korea Institute of Convergence Signal Processing
• /
• 2003.06a
• /
• pp.296-299
• /
• 2003

This paper shows a novel and simple IIR/FIR QMF(quadrature mirror filter) filter banks, mixed IIR and FIR structure. Here, FIR filters used for phase compensation. In this paper, we introduced analysis and synthesis filter banks, which used FIR linear phase filters and all pass filters. In result, phase response of analysis and synthesis filter banks become approximately linear characteristic. Simultaneously, a liasing distortion can be completely canceled.

• Proceedings of the KAIS Fall Conference
• /
• 2010.05a
• /
• pp.273-276
• /
• 2010

음성신호를 관찰하였을 때 성문특성으로 인해서 고주파쪽 특성이 약화되는 경향이 있다. 약화된 고주파 특성을 보상하기 위하여 프리 엠퍼시스 필터를 통해 보상하고 있다. 프리 엠퍼시스 필터를 간단한 수식으로 표현하면 y(n)=s(n)-As(n-1)와 같이 차분 방정식으로 나타낼 수 있다. 여기서 A값은 보통 0.9에서 1사이의 값을 사용한다. 본 논문에서는 QMF 필터를 이용하여 입력신호를 고주파와 저주파의 2개의 대역으로 분할하여 각 밴드에 프리 엠퍼시스 필터를 적용하여 약화되어진 특성을 정확히 보상하는 방법을 제안한다.

• Proceedings of the IEEK Conference
• /
• 1998.10a
• /
• pp.1179-1182
• /
• 1998

광대역 능동잡음제거는 수 백개의 적응필터 탭 수를 갖는다. 탭 수가 긴 적응필터는 많은 계산량이 요구된다. 본 논문에서는 적응계수 벡터가 서브밴드로 계산되는 M-채널 QMF 필터뱅크와 적응필터를 이용한 자동차 소음제거 모델을 제안한다. 분해 필터뱅크와 합성 필터뱅크는 cosine-modulated pseudo QMF 필터를 사용한다. 오차경로의 전달특성을 온라인 인식하기 위한 기준신호는 적응필터의 출력신호와 저주파대역의 서브밴드 출력신호와의 차신호를 사용한다. 따라서 제안한 자동차 소음제거 모델은 계산량이 적고 수렴속도가 빠른 견실한 시스템이 되도록 제안한다.