In this paper in order to reduce the encoding complexity required in the full search vector quantization(VQ), a new classified vector quantization(CVQ) technique is described employing the minimum-distance classifier. The determination of the optimal subcodebook sizes for each class is an important task in CVQ designs and is not an easy work. Therefore letting the subcodebook sizes be equal. A CVQ technique. Which satisties the optimal CVQ condition approximately, is proposed. The proposed CVQ is a kind of the partial search VQ because it requires a search process within each subcodebook only, and the minimum encoding complexity since the subcodebook sizes are the same in each class. But simulation results reveal while the encoding complexity is only O(N$^{1/2}$) comparing with O(N) of the full-search VQ. A simple systolic array, which has the through-put of k, is also proposed for the implementation of the VQ. Since the operation of the classifier is identical with that of the VQ, the proposed array is applied to both the classifier and the VQ in the proposed CVQ, which shows the usefulness of the proposed CVQ.

This paper presents a fast massively parallel automatic test pattern generator for digital combinational logic circuits using neural networks. Automatic test pattern generation neural network(ATPGNN) evolves its state to a stable local minima by exchanging messages among neural network modules. In preprocessing phase, we calculate the essential assignments for the stuck-at faults in fault list by adopting dominator concept. It makes more neurons be fixed and the system speed up. Consequently. fast test pattern generation is achieved. Test patterns for stuck-open faults are generated through getting initialization patterns for the obtained stuck-at faults in the corresponding ATPGNN.

In this paper, we propose a multimedia query language MQL which defines and manipulates multimedia data as integration of monomedia data in time and space. The MQL is designed for a multimedia data model, called the object-relationship model, and based on the multimedia object calculus which formally describes operations on multimedia data. The SQL- like syntax for class definition and object manipulation, such as retrieval, insert, update, and delete, is defined. We show how the MQL can represent the user queries using composite temporal-spatial class structures and various relationships, such as equivalence and sequence.

Many kinds of large scale multiprocessing and parallel-processing systems have recently been developed. The contention on the shared data caused by multiple processors may degrade system performance. So, processor synchronization has become one of the important issues in these systems. To solve the synchornization issues, a lot of software and hardware schemes based on spin lock have been proposed. Although software schemes are easy to implement, hardware schemes are preferred in many systems to gain optimized performance. This paper proposes an efficient processor synchronization scheme, called QCX,and describes its design considerations, hardware, algorithm, protocol. Also, in this paper, the performance of QCX has been evaluated with QOLB[5] and LBP[7] using a simulation. The simulation, with varying the number of processor and the contention on shared variables, measured the average execution times of a workload. The simulation results show that the performances of QCX is best when practicability is considered. QCX is more efficient than QOLB and LBP in two aspects. First, the hardware of QCX is more simple and cost-effective because the cache structure need not be changed. Secondly, QCX is more general because it uses a generic atomic instruction.

In most application programs, loops usually comprise most of the computation in a program and are the most important source of parallelism. When loops are executed on multiprocessors, the cross iteration data dependences need to be enforced by synchronization between processors. Existing synchronization schemes have been studied mainly on the loop with constant dependence distance. When these schemes are applied to the loop with variable dependence distance, there exists lots of overhead by the use of unnecessary synchronization variables and execution of unuseful synchronization instructions. Even though there exist various variable synchronization schemes, they have a lot of run-time overhead to compute synchronization information. In this paper, we present a new synchronization scheme, Synch-Free/Synch-Hold for managing synchronization efficiently on the loop with constant and variable dependence distance.

Conditional branch instructions are a major obstacle to the increasing of RISC processor performance, because they can break the smooth flow of instructions; the issuing of instructions after a branch instruction must often wait until the condition is resolved. This paper proposes a hardware scheme which has a duplicated fetching logic to reduce the penalty imposed by conditional branch instructions. The proposed shceme has a buffer to maintain states of processor, which supports the precise interrupt. We make use of two code segments to test the performance and their results were compared with those of the delayed branch. We got the result that the proposed scheme reduces the branch penalty extremely.

In LD driver with HBT's. the speed characteristics of HBT's are deteriorated very much mainly due to the source resistance(Rs) of the signal applied to the LD driving HBT. We improved the bandwidth of LD driver by 2-5GHz with modifications of EF buffer. Because the modified buffers are composed of EF structure, their bandwidths are decreased rapidly as Rs is increased. When Rs is large these buffers decrease the bandwidth of the LD driver rather than increase it. To solve this problem, we propose a new buffer structure which contaings new charging and discharging path for the parasitic collector capacitance of the HBT. For Rs>100${\Omega}$, it shows superior characteristics of improving bandwidth to the EF buffers. It also shows good gain characteristics.

We proposed a XMESH(crossed-mesh) topology as a suitable interconnection for the massively parallel computer architectures, and presented performance analysis of the proposed interconnection topology. Horizontally, the XMESH has the same links as those of the toroidal mesh(TMESH) or toroid, but vertically, it has diagonal cross links instead of the vertical links. It reveals desirable interconnection characteristics for the massively parallel computers as the number of nodes increases, while retaining the same structural advantages of the TMESH such as the symmetric structure, periodic placement of subsystems, and constant degree, which are highly recommended features for VLSI/WSI implementations. Furthermore, n*k XMESH can be easily expanded without increasing the diameter as long as n.leq.k.leq.n+4. Analytical performance evaluations show that the XMESH has a shorter diameter, a shorter mean internode distance, and a higher message completion rate than the TMESH or the diagonal mesh(DMESH). To confirm these results, an optimal self-routing algorithm for the proposed topology is developed and is used to simulate the average delay, the maximum delay, and the throughput in the presence of contention. In all cases, the XMESH is shown to outperform the TMESH and the DMESH regardless of the communication load conditions or the number of nodes of the networks, and can provide an attractive alternative to those networks in implementing massively parallel computers.

This paper presents the design of a CMOS programmable slew rate operational amplifier based upon a newly proposed concept, that is a switched parallel current subtraction circuit with adaptive biasing technique. By utilizing the newly designed circuit, it was proven that slew rate was linearly controlled and power dissipation was optimized. If the programmable slew rate amplifier is employed into mixed signal system, it can furnish the convenience of timing control and optimized power dissipation. Simulated data showed the slew rate ranging from 5. 83V/$\mu$s to 41.4V/$\mu$s, power dissipation ranging from 1.13mW to 4.1mW, and the other circuit performance parameters were proven to be comparable with those of a conventional operational amplifier.

When codebooks designed by a clustering algorithm using training sets, a time-averaged distortion, which is called the inside-training-set- distortion (ITSD), is usually calculated in each iteration of the algorithm, since the input probability function is unknown in general. The algorithm stops if the ITSD no more significantly decreases. Then, in order to test the trained codebook, the outside-training-set-distortion (OTSD) is to be calculated by a time-averaged approximation using the test set. Hence codebooks that yield small values of the OTSD are regarded as good codebooks. In other words, the calculation of the OTSD is a criterion to testing a trained codebook. But, such an argument is not always true if some conditions are not satisfied. Moreover, in order to obtain an approximation of the OTSD using the test set, it is known that a large test set is requared in general. But, large test set causes heavy calculation com0plexity. In this paper, from the analyses in [16], it has been revealed that the enough size of the test set is only the same as that of the codebook when codebook size is large. Then a simple method to testing trained codebooks is addressed. Experimental results on synthetic data and real images supporting the analysis are also provided and discussed.

In this paper, we proposed MLP(MultiLayer Perceptron) based HMM's(Hidden Markov Models) with bounded state duration for isolated word recognition. The minimum and maximum state duration for each state of a HMM are estimated during the training phase and used as parameters of constraining state transition in a recognition phase. The procedure for estimating these parameters and the recognition algorithm using the proposed HMM's are also described. Speaker independent isolated word recognition experiments using a vocabulary of 10 city names and 11 digits indicate that recognition rate can be improved by adjusting the minimum state durations.

This paper describes a speaker adaptation method to improve the recognition performance of MLP(multiLayer Perceptron) based HMM(Hidden Markov Model) speech recognizer. In this method, we use lst-order linear transformation network to fit data of a new speaker to the MLP. Transformation parameters are adjusted by back-propagating classification error to the transformation network while leaving the MLP classifier fixed. The recognition system is based on semicontinuous HMM's which use the MLP as a fuzzy vector quantizer. The experimental results show that rapid speaker adaptation resulting in high recognition performance can be accomplished by this method. Namely, for supervised adaptation, the error rate is signifecantly reduced from 9.2% for the baseline system to 5.6% after speaker adaptation. And for unsupervised adaptation, the error rate is reduced to 5.1%, without any information from new speakers.

This paper presents a new method for automatically calibrating robot link (Kinematic) parameters during the process of estimating motion parameters of a moving object. The motion estimation is performed based on stereo cameras mounted on the end-effector of a robot manipulator. This approach significantly differs from other calibration approaches in that the calibration is achieved by simply observing the motion of the moving object (without resorting to any other external calibrating tools) at numerous and widely varying joint-angle configurations. A differential error model, which expresses the measurement errors of a robot in terms of robot link parameter errors and motion parameters, is developed. And then a measurement equation representing the true measurement values is derived. By estimating the above two kinds of parameters minimizing the difference between the measurement equations and the true moving pattern, the calibration of the robot link parameters and the estimation of the motion parameters are accomplished at the same time.

In this paper, a variable time-scale modification using SOLA(Synchronized OverLap and Add) is proposed, which takes into consideration the different time-scaled characteristics of voiced and unvoiced speech, Generally, voiced speech is subject to higher variations in length during time-scale modification than unvoiced speech, but the conventional method performs time-scale modification at a uniform rate for all speech. For this purpose, voiced and unvoiced speech duration at various talking speeds were statistically analyzed. The sentences were then spoken at rates of 0.7, 1.3, 1.5 and 1.8 times normal speed. A clipping autocorrelation function was applied to each analysis frame to determine voiced and unvoiced speech to obtain respective variation rates. The results were used to perform variable time-scale modification to produce sentences at rates of 0.7, 1.3, 1.5, 1.8 times normal speed. To evaluate performance, a MOS test was conducted to compare the proposed voiced/unvoiced variable time-scale modification and the uniform SOLA method. Results indicate that the proposed method produces sentence quality superior to that of the conventional method.

In this study, radar data integrating and processing systems were designed for the data processing of various information from many kinds of radar in a single data processing system. The characteristics of the data integrating system were analyzed by the system simulation with the queueing theory. The designed data integrating systems can be divided into a centralized and a distributed type. In the system structure, we used UNIX message que as the real time processor and the queueing theory for the performance evaluation of the information flow in the systems. For the analysis of the performance of inforamtion flow in both models, queueing theory was applied to and implemented with the simulation package, OPNET system and C language. From the simulation result we could understand the system factors which effect the system performance and characteristics on the data processing.

This paper proposes an approach of identifying nonlinear mappings from input/output data. The approach is based on the universal approximation by the fuzzy integration of local affine mappings. A connectionist model realizing the universal approximator is suggested by using a processing unit based on both the radial basis function and the weighted sum scheme. In addition, a learning method with self-organizing capability is proposed for the identifying of nonlinear mapping relationships with the given input/output data. To show the effectiveness of our approach, the proposed model is applied to the function approximation and the prediction of Mackey-Glass chaotic time series, and the performances are compared with other approaches.