Parallel VLSI Implementation Research Articles

Efficient MDCT Recursive Structure for VLSI Implementation

Modified discrete cosine transform (MDCT) is a computationally intensive fundamental processing block for mapping time into frequency in various versions of MPEG audio coding. An efficient algorithm for MDCT computation is a basic requirement to realize low-cost audio codec. This paper introduces a new recursive algorithm and architecture for MDCT. The proposed architecture requires minimum hardware and execution time. Folding of the input sequence is performed multiple times in the subsequent groups which results in a significant reduction in real multiplication and real addition operations. Also, fewer number of execution cycles are required to generate the output through this algorithm. The proposed structure is suitable for parallel VLSI implementation as it is simple, regular and modular.

Novel MIMO Detection Algorithm for High-Order Constellations in the Complex Domain

A novel detection algorithm with an efficient VLSI architecture featuring efficient operation over infinite complex lattices is proposed. The proposed design results in the highest throughput, the lowest latency, and the lowest energy compared to the complex-domain VLSI implementations to date. The main innovations are a novel complex-domain means of expanding/visiting the intermediate nodes of the search tree on demand, rather than exhaustively, as well as a new distributed sorting scheme to keep track of the best candidates at each search phase. Its support of unbounded infinite lattice decoding distinguishes the present method from previous K-Best strategies and also allows its complexity to scale sublinearly with the modulation order. Since the expansion and sorting cores are data-driven, the architecture is well suited for a pipelined parallel VLSI implementation. The proposed algorithm is used to fabricate a 4×4, 64-QAM complex multiple-input-multiple-output detector in a 0.13-μm CMOS technology, achieving a clock rate of 417 MHz with the core area of 340 kgates. The chip test results prove that the fabricated design can sustain a throughput of 1 Gb/s with energy efficiency of 110 pJ/bit, the best numbers reported to date.

Recursive computation of Tchebichef moment and its inverse transform

Tchebichef moment is a novel set of orthogonal moment applied in the fields of image analysis and pattern recognition. Less work has been made for the computation of Tchebichef moment and its inverse moment transform. In this paper, both a direct recursive algorithm and a compact algorithm are developed for the computation of Tchebichef moment. The effective recursive algorithm for inverse Tchebichef moment transform is also presented. Clenshaw's recurrence formula was used in this paper to transform kernels of the forward and inverse Tchebichef moment transform. There is no need for the proposed algorithms to compute the Tchebichef polynomial values. The approaches presented are more efficient compared with the straightforward methods, and particularly suitable for parallel VLSI implementation due to their regular and simple filter structures.

Adaptive Neural Network for Nuclear Medicine Image Restoration

A novel adaptive neural network is proposed for image restoration using a nuclear medicine gamma camera based on the point spread function of measured system. The objective is to restore image degradation due to photon scattering and collimator photon penetration with the gamma camera and allow improved quantitative external measurements of radionuclidesin-vivo . The specific clinical model proposed is the imaging of bremsstrahlung radiation using ^32P and^90 Y because of the enhanced image degradation effects of photon scattering, photon penetration and poor signal/noise ratio in measurements of this type with the gamma camera. This algorithm model avoids the common inverse problem associated with other image restoration filters such as the Wiener filter. The relative performance of the adaptive NN for image restoration is compared to a previously reported order statistic neural network hybrid (OSNNH) filter by these investigators, a traditional Weiner filter and a modified Hopfield neural network using simulated degraded images with different noise levels. Quantitative metrics such as the change of signal to noise ratio (ΔSNR) are used to compare filter performance. The adaptive NN yields comparable results for image restoration with a slightly better performance for the images with higher noise level as often encountered in bremsstrahlung detection with the gamma camera. Experimental attenuation measurements were also performed in a water tank using two radionuclides, ^32P and ^90 Y, typically used for antibody therapy. Similar values for an effective attenuation coefficient was observed for the restored images using the OSNNH filters and adaptive NN which demonstrate that the restoration filters preserves the total counts in the image as required for quantitative in-vivo measurements. The adaptive NN was computationally more efficient by a factor 4–6 compared to the OSNNH filter. The filter architecture, in turn, is also optimum for parallel processing or VLSI implementation as required for planar and particularly for tomographic mode of detection using the gamma camera. The proposed adaptive NN method should also prove to be useful for quantitative imaging of single photon emitters for other nuclear medicine tomographic imaging applications using positron emitters and direct X-ray photon detection.

Completely Parallel Realizations for Multidimensional Digital Filters

This paper proposes a new method for realizing the general form multi-dimensional (M-D) digital filter as the combination of parallel sections consisting of cascaded one-dimensional (1-D) digital filters. First, we use the coefficients of an M-D filter to construct an M-D coefficient array; then we propose a technique for decomposing the M-D coefficient array into the summation of the outer products of real-valued vectors. Since the real-valued vectors represent the coefficients of 1-D filters with different delay elements, the decomposition of the M-D coefficient array implies that a general form M-D filter can be decomposed into the combination of completely parallel sections, and each of them consists of cascaded 1-D filters. Consequently, the structure of the realized M-D filter possesses high parallelism, regularity, and modularity, and it is suitable for high-speed parallel signal processing and VLSI implementation.

Fast algorithm for computing discrete cosine transform

An efficient method for computing the discrete cosine transform (DCT) is proposed. Based on direct decomposition of the DCT, the recursive properties of the DCT for an even length input sequence is derived, which is a generalization of the radix 2 DCT algorithm. Based on the recursive property, a new DCT algorithm for an even length sequence is obtained. The proposed algorithm is very structural and requires fewer computations when compared with others. The regular structure of the proposed algorithm is suitable for fast parallel algorithm and VLSI implementation.

A restoration algorithm for P-32 and Y-90 bremsstrahlung emission nuclear imaging: a wavelet-neural network approach.

A novel wavelet-based neural network (WNN) filter is proposed for image restoration as required for imaging of beta emitters by bremsstrahlung detection using a gamma camera. Quantitative imaging of beta emitters is important for the in vivo management of antibody therapy using either P-32 or Y-90. The theoretical basis for the general case for M-channel multiresolution wavelet decomposition of the nuclear image into different subimages is developed with the objective of isolating the signal from noise. A modified Hopfield neural network (NN) architecture is then used for multichannel image restoration using the dominant signal subimages. The NN model avoids the common inverse problem associated with other image restoration filters such as the Wiener filter. The relative performance of the WNN for image restoration, for M = 2 channel, is compared to a previously reported order statistic neural network hybrid (OSNNH) filter. Initially simulated degraded images of known structures with different noise levels are used. Quantitative metrics such as the normalized mean square error (NMSE) and signal-to-noise ratio (SNR) are used to compare filter performance. The WNN yields comparable results for image restoration with suggested slightly better performance for the images with higher noise levels as often encountered in bremsstrahlung detection. Attenuation measurements were performed using two radionuclides, 32P and 90Y as required for calibration of the gamma camera for quantitative measurements. Similar values for an effective attenuation coefficient were observed for the restored images using the OSNNH filters (32P: mu = 0.122 cm-1, 90Y: mu = 0.135 cm-1) and WNN (32P: mu = 0.122 cm-1, 90Y: mu = 0.135 cm-1) filters with slightly higher values obtained for the raw data (32P: mu = 0.142 cm-1, 90Y: mu = 0.142 cm-1) for a 3.5-cm source size. The WNN, however, was computationally more efficient by a factor of 4 to 6 compared to the OSNNH filter. The filter architecture, in turn, is also optimum for parallel processing or VLSI implementation as required for planar and particularly for SPECT mode of detection.

3-dimensional systolic architecture for parallel VLSI implementation of the discrete cosine transform

Two different linear systolic arrays have been suggested for the computation of discrete cosine transform (DCT). The proposed linear arrays are complementary to each other in the sense that the output of the linear arrays of one type may be fed as the input for the linear arrays of the other type. This feature of the proposed linear arrays has been utilised for designing a bilayer structure for computing the prime-factor DCT. It is interesting to note that the proposed structure does not require any hardware/time for transposition of the intermediate results. The desired transposition is achieved by orthogonal alignment of the linear arrays of the upper layer with respect to those of the lower layer. The proposed structures provide high throughput of computation due to fully pipelined processing, and massive parallelism employed in the bilayer architecture.

Visual inspection in industrial manufacturing

Visual quality control is a demanding task of increasing importance in industrial manufacturing. Both speed and flexibility are of paramount importance for viable and competitive inspection systems. We have developed a dedicated neural network architecture for anomaly detection that can easily be trained by a single presentation of examples and is amenable to massively parallel VLSI implementation. We focus here on our ASIC and prototype system design effort for this network. >

Parallel regional projection transformation (RPT) and VLSI implementation

A new regional projection transformation (RPT) to recognize unconnected patterns and patterns with isolated noise is presented. This new approach simplifies the process of recognizing compound patterns by transforming them into an integral object. Two kinds of RPT transforms are described and analysed: (1) diagonal-diagonal regional projection transformation (DDRPT), and (2) horizontal-vertical regional projection transformation (HVRPT). The patterns transformed by these two methods have several important properties which can simplify contour processing. The essential parallelisms in DDRPT and HVRPT can also facilitate their parallel implementation and parallel algorithms for both DDRPT and HVRPT, and their VLSI implementation are also designed. They can speed up the recognition process considerably with a time complexity of O(N) for processing a pattern with size of N × N, compared with O( N 2) using a uniprocessor.

Parallel VLSI matrix pencil algorithm for high resolution direction finding

Algorithms to find the directions of arrival (DOAs) of multiple signals from measurements on an array of antenna doublets (ESPRIT method) and their parallel VLSI implementation are discussed. In particular, algorithms that allow large-scale pipelining and use only robust, unitary transformations are identified. This problem is solved by a matrix pencil approach in which the generalized eigenvalues of a pair of data matrices are determined. A modified Stewart Jacobi approach is used for which convergence is improved and parameter computations are simplified. The resulting architecture is a two-layer Jacobi array that can handle all the subproblems: two QR factorizations, two SVDs, and a single generalized Schur decomposition. The mapping of the subproblems on a single parallel array of CORDIC processors is considered.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Error diffusion coding for A/D conversion

Various novel techniques for A/D conversion of signals subject to a fidelity criterion are presented, leading to optimum digital representations, in which each signal sample is not necessarily quantized to the closest reconstruction level. Quantization is treated as an optimization problem, and the tradeoffs among sampling rate, quantization stepsize, and quantization distortion are examined. It is shown that symmetric neural networks offer a natural means for efficient implementation of the proposed technique. Applications include digital image halftoning, as well as all forms of PCM coding and oversampled A/D conversion. It is shown that concepts and structures used in digital image halftoning are directly applicable to oversampled sigma-delta modulation of sound signals. A novel kind of parallel analog network is introduced and shown to be appropriate for this task. These networks contain a nonmonotonic nonlinearity in lieu of the sigmoid function and perform error diffusion in all directions. Ideas for massively parallel analog VLSI implementation are offered.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Parallel VLSI Implementation of the Kalrman Filter

The problem of parallel implementation of the square-root Kalman filters is addressed. At the system level, our approach is to apply systolic-type VLSI processor arrays as basic building blocks to accelerate the matrix operations required in each iteration. To maximize the parallelism, we also exploit an inter-array pipelining scheme through the overlapping of execution between successive processor arrays. We estimate that with (5n2 + r2 + 8nr + n + 3r)/2 processors, it would take max[(4n + 2r, 2n + 4r-2)] time units to complete one Kalman filter iteration, where n is the dimension of the underlying state space model and r is the dimension of the input vector.

VLSI based design principles for MIMD multiprocessor computers with distributed memory management

Design principles for MIMD multiprocessor computers with virtual memory based on a common, global and uniform logical address space, supporting parallel, procedural languages such as Ada (Ada is a registered trademark of the US Government, AJPO), are discussed. The major design issues are identified and suggested solutions given, the most important of which are distributed, associative address translation, and local mechanisms supporting efficient resource allocation policies to reduce over-all communication costs. Arguments are given for using shared memory and bus-based global communication. Some preliminary studies of bus-based intercommunication schemes, parallel language implementation, capacity simulation and VLSI implementation are reviewed, as well as a number of existing experimental and commercial multiprocessors. Finally an experimental system for evaluation of different mechanisms and policies in systems of the suggested type is outlined.