Search

In this paper, we propose a design method of perfect reconstruction (PR) nonuniform filter banks. Filter banks are roughly separated into nonparaunitary and paraunitary ones. Paraunitary filter banks have some advantages: for example, the frequency responses of synthesis filters are identical to those of analysis filters. However, the severe constraint makes it difficult to design them and, especially in the nonuniform case, the design method becomes complex for some frequency-division cases. Thus, we concentrate on nonparaunitary filter banks and present a design method for nonuniform filter banks with arbitrary lengths and frequency divisions. © 1997 Scripta Technica, Inc. Electron Comm Jpn Pt 3, 80(10): 51–59 1997 [ABSTRACT FROM AUTHOR]

This paper describes an efficient structure for generalized lapped orthogonal transforms (GenLOT) for finite-duration sequences, which we expect to be used for subband image coding. When linear convolution is carried out on the finite-duration sequences and filter, the number of points in the output signal increases. Therefore, the linear convolution is not desirable in subband image coding. In order to solve this problem, the convolution is carried out after the finite-duration sequences are symmetrically and periodically extended (symmetric extension method). This method is superior to circular convolution because the former method causes less distortion at the signal boundaries. In the GenLOT structure proposed in this paper, the symmetric extension can be realized without redundant operations due to signal extension and, furthermore, with fast implementation. This structure is considered to be a generalization of the lapped orthogonal transform (LOT) for finite-duration sequences. In addition, its application to the M-band discrete-time wavelet transform (DTWT) will be described. Finally, it will be shown that an efficient structure for finite-duration sequences can be realized in the M-band DTWT. © 1997 Scripta Technica, Inc. Electron Comm Jpn Pt 3, 80(10): 39–50, 1997 [ABSTRACT FROM AUTHOR]

There are some applications that require oversampled filter banks, for example, adaptive signal processing using multirate filter banks. Little study was done, however, for such filter banks compared with maximally decimated ones. In this paper, we propose a design method for oversampled filter banks which satisfy the perfect reconstruction and linear phase properties. Moreover, we consider the design method for both orthogonal and biorthogonal ones. When the decimation ratio is not a divisor of the number of channels, we find it impossible for analysis and synthesis filters to have the same frequency responses by considering the mechanism of aliasing cancellation. In other words, biorthogonal filter banks can be designed under an arbitrary decimation ratio, while it is restricted to a divisor of the number of channels in the orthogonal case. We first derive the conditions for perfect reconstruction and linear phase for both biorthogonal and orthogonal ones. Then, we present the design method for both cases. Finally, some design examples are shown. © 1997 Scripta Technica, Inc. Electron Comm Jpn Pt 3, 80(10): 78–86, 1997 [ABSTRACT FROM AUTHOR]

The filter on monolithic LSI for communication and measurement is realized by the digital signal processor (DSP) or the switched capacitor (SC) circuit. In the analog-digital interface, high-speed processing, applications requiring low power consumption operations, and low cost, SC circuits will be superior to the DSP circuits. The authors devised the SC circuit called parallel cyclic-type, which can further enhance the high-frequency and low power-consumption operation. The single-rate FIR and IIR filters have been proposed using the circuit. In this paper, the high frequency and low power-consumption FIR and IIR decimation filters using the SC parallel cyclic-type circuits are proposed. The effects of the finite gain bandwidth (GB product) of the operational amplifier and of the parasitic capacitance on the filter characteristics are analyzed. The proposed filters and the decimation filter by Franca et al. are constructed using discrete elements and the characteristics are examined. The characteristics are compared with the computer simulations. It is verified as a result that the proposed decimation filter has a much better performance than the filter by Franca et al. [ABSTRACT FROM AUTHOR]

Control and protection equipment in power systems requires higher sensitivity and operational reliability to meet today's changing power system requirements. The voltage-measuring deviation requirement for advanced voltage and var control equipment is less than 0.1 percent under conditions of harmonic distortion in the voltage waveform and power-system frequency variation. Studies on digital signal processing suitable for electric power systems showed that these requirements are satisfied using fast sampling and very fast 32-bit floating point operations by a Digital Signal Process (DSP). This paper describes the design philosophy of a high-precision power system voltage-measuring method using fast sampled data. In addition, total voltage-measuring deviation characteristics under a combination of the techniques are described along with digital filter characteristics, frequency-measuring deviation characteristics, frequency variation versus gain-compensation characteristics of a digital filter, and peak value operating principles. [ABSTRACT FROM AUTHOR]