Recently, some speech recognition methods using fusion of visual and auditory information have been researched. In this paper, a study on the mouth shape image suitable for fusion of visual and auditory information has been described. Features of mouth shape which are extracted from gray level image and binary image are adopted, and speech recognition using linear combination method has been performed. From results of speech recognition, the studies on the mouth shape features which are effective in fusion of visual and auditory information have been performed. And the effectiveness of using two kinds of mouth shape features also has been confirmed.

Analyzing multipath propagation structure is important to develop anti-fading techniques for high-speed digital radio systems. Several techniques have been employed to measure delay profiles and/or arrival angles. This paper presents a simultaneous estimation method of delay times and arrival angles of indoor multipath waves. We obtain frequency-domain data at different receiving antenna positions using a network analyzer. We estimate the propagation parameters by means of a two-dimensional MUSIC algorithm. In order to obtain reliable results, a two-dimensional discrete inverse Fourier transform and a gating technique are employed before the MUSIC algorithm. Simulation and experimental results show that the proposed method can estimate the propagation parameters properly.

As Object-Oriented Database Systems (OODBS) play an increasingly important role in advanced database systems, OODBS performance becomes a significant issue. It is well known that there is a strong relationship between performance and the concurrency control algorithms employed by the Database Management System (DBMS). Class Granularity Locking (CGL) and Class Hierarchy Granularity Locking (CHGL) are proposed as the concurrency control algorithms for OODBS to minimize the locking overhead. However, their basic characteristics, including the licking overhead and concurrency, have not been extensively investigated and it is not known which one is most appropriate for the general case. In this paper, we construct a simulation model for OODBS and carry out several performance evaluation studies on these two Class-Hierarchy Locking protocols and the Non Class-Hierarchy Locking (NCL) protocol. The NCL protocal is a variation of the conventional two phase locking protocol being applied to OODBS data structures.

Queries to generate a map from geographic detabases are too large and complex for users to specify all components in it. Thematic parts of the map should be described by users as ad hoc queries. However, background parts of it should be inferred from users' queries corresponding to the thematic parts. Furthermore, it is important for the map systems to lead users' constructing their ad hoc queries and to infer visualization methods applied to the data retrieved by the ad hoc queries. This paper discusses a framework to infer supplemental queries and visualization methods in order to make the retrieval results into a feasible map using geographic domain hierarchical levels, geographic domain thesauruses and existing example queries. The framework allows users to know mismatches of components in queries, inappropriate queries for maps, and deriving candidates for additional components in queries.

Due to the large and complex information processing systems, formal description methods are needed for specification of systems and their efficient and reliable designs. During the early stage of system design, it is often necessary to modify or change system requirements which may influence the whole system design. We have proposed a new flexible description methodology, which copes with the modifications or changes in the system requirements, in order to obtain the formal specification of the system. We have also shown that function requirements can be modeled by a Logical Petri Net (LPN), which is a kind of extended Petri Nets, in order to derive the formal specification. In this paper, we propose a verification method of system requirements that contain some kinds of logical errors. Further, we show a method to decompose and refine a requirement description hierarchically, and discuss how to derive a formal specification from a requirement description flexibly along our refinement method against the changes of the requirement description in the system.

Efficient retrieval of nested objects is an important issue in advanced database systems. So far, a number of indexing methods for nested objects have been proposed. However, they do not consider retrieval of nested objects based on the set comparison operators such as and . Previouly, we proposed four set access facilities for nested objects and compared their performance in terms of retrieval cost, storage cost, and update cost. In this paper, we extend the study and present refined algorithms and cost formulas applicable to more generalized situations. Our cost models and analysis not only contribute to the study of set-valued retrieval but also to cost estimation of various indexing methods for nested objects in general.

An automatic transistor-level performance fault tracing method is proposed which is applicable to the case where only CAD layout data is available in the CAD-linked electron beam test system. The technique uses an integrated algorithm that combines a previously proposed transistor-level fault tracing algorithm and a successive circuit extraction from CAD layout data. An expansion of the algorithm to the fault tracing in a combined focused ion beam and electron beam test system which enables us to measure signals on the interconnections in the lower layers is also described. An application of the technique to a CMOS model layout with about 100 transistors shows its validity.

This paper is concerned with synthesizing VLSI array processors from iterative algorithms. Our primary objective is to obtain the highest processor efficiency but not the shortest completion time. Unlike most of the previous work that assumes the index space of the given iterative algorithm to be boundless, the proposed method takes into account the effects of the boundaries of the index space. Due to this consideration, the pseudo-dependence relations are excluded, and most of the independent computations can therefore be uniformly grouped. With the method described in this paper, the index space is partitioned into equal-size blocks and the corresponding computations are systematically and uniformly mapped into processing elements. The synthesized VLSI array processors possess the attractive feature of very high processor efficiency, which, in general, is superior to what is derived from the conventional linear transformation methods.

This paper proposes a masked data transferring method for the write-back controlled disk cache system employing a fixed-length recording disk drive, enabling data transfer of discontinuous sectors on the same track between the cache and the disk. This paper also evaluates the method. In write-back controlled disk cache sytems, random write requests cause dirty data (write-pending data on a cache) on discontinuous areas on the cache. It is likely that several sectors on the same track become dirty. These dirty sectors must be written onto the disk according to the cache management scheme. In conventional data transferring methods between a disk cache and a disk drive, plural sectors can be transferred in one single operation when the sectors are adjacent, but discrete sectors must be transferred by individual operations. In the methods, an address of the head sector and number of sectors to be transferred are given to the transfer unit. For example, when two sectors on the same track are located closely but not adjacently, and data transfer is requested for those two sectors, the transfer operation for the second sector must be prepared after the first transfer had completed and before the second sector arrives under the disk head. Although the time for the head to pass by the uninterested sector is often too short for the software overhead for the first transfer to be completed and the second transfer to be prepared, which leads to an unwanted extra rotation of the disk. With the masked transferring method proposed in this paper, the micro program creates a bit-map specifying the target sectors to be transferred and passes it to the data transfer unit, enabling to transfer the discontinuous sectors without latency. The method was evaluated using OLTP warkloads. Results show an improvement in random I/O throughput of between 8% and 27%. The masked transferring method is adopted in Hitachi's A-6521 disk subsytems, shipped since December 1993.

Information capture is a very desirable and important function in portable computers. The "Multimedia Notebook" is a portable tool for capturing information in multimedia format, which includes photos, voice, and handwritten memos. Recent portable units, sometimes called PDA (Personal Digital Assistants) or Communicators, have PIM (Personal Information Manager) software and some communication facility. Their purpose is to enlarge the desktop environment to follow the user outside the office. This is one application of portable equipment, however we felt that hand-held equipment can also be used for more general information capture. In the past, information capture was limited, because people had to carry bulky equipment to the information source. Recent portable computers that have the capability to handle still and motion pictures, voice, and handwritten drawings allow the implementation of more information capture capabilities. Capture of handwritten notes has already been implemented on portable equipment like PDAs. However, this application doesn't make the most of its potential. We feel that the Multimedia Notebook should integrate handwritten memos with pictures and voice. The advantage is that users can capture what they watch and hear easily in a variety of media for later review. The information in each medium complements the others. When arranging the recorded information, it is easy to use each medium efficiently. We have examined the human interface and designed it for user-friendliness and to be comparable to pen and paper. The prototypes also have a capture buffer which can operate continuously to capture voice and pictures that would otherwise be lost because of the user's delay in starting to record.

Active net is a deformable model which utilizes the network analogy of a physical region. In the model, the region of a target is detected by minimizing the energy defined for the sample points of the model. The region of the target is extracted using fixed network topology in the orginally proposed algorithm. In this paper, we introduce the network reconfiguration mechanisms such as tearing and division to realize multiple objects detection and complex object detecion. The introduced algorithm dynamically unlinks the arcs of the network when their strain value exceeds predefined threshold level. In the method, we propose a new image energy which improves the position sensitivity of edges without increasing computation cost. Experimental results for images taken by video camera show the validity of our approach.

This paper establishes a general relation between the two-dimensional Least Mean Square (2-D LMS) algorithm and 2-D discrete orthogonal transforms. It is shown that the 2-D LMS algorithm can be used to compute the forward as well as the inverse 2-D orthogonal transforms in general for any input by suitable choice of the adaptation speed. Simulations are presented to verify the general relationship results.

A photonic/video hybrid system for optical information processing by synthesis of the coherence function is proposed. Optical coherence function can be synthesized to have delta-function-like shape or notch shape by using direct frequency modulation of a laser diode with an appropriate waveform. Therefore, by choosing only the interference component in the interferometer, information processing functions can be obtained. The photonic/video hybrid system proposed provides a novel way to choose the interference component, which can improve the spatial resolution compared with our previous system with holographic technique. Selective extraction two-dimensional (2-D) information from a three-dimensional (3-D) object is successfully performed in basic experiments.

Information systems to support cooperative work among people should be first designed to help humam communication. However, there are few systems based on the analysis of human communication. Standing on this situation, we propose a meeting support system for the participants' understandings by indicating the suitable information about the topic of the scene". Our system provides only useful information by monitoring each statement without complex methods. To show something useful multi-media information for members, we propose the following structure of the meeting on the basis of the analysis of communication. Each statement is classified into two levels, either; a statement about the progress" of the meeting (context-level utterances) or, a statement about objects" (content-level utterances). Further, content-level utterances are classified into two types, position utterances and argument utterances. Using this classification of statements, the proceeding of the meeting is represented as the tree model which is called a context-tree". If the structure of meetings is fixed, it is possible to select only useful information from all shared information for members by analyzing each content-level utterance. The system introduced in this paper shows appropriate multi-media information about the topic of the scene" by using the above model. We have implemented a prototype system based on the above ideas. Moreover, we have mode some experiments to show the effectiveness of this system. Those results show that our method is effective to improve the productivity" of meetings.

In this paper face feature detection and tracking are discussed, using methods called edge pixel counting and deformable circular template matching. Instead of utilizing color or gray scale information of the facial image, the proposed edge pixel counting method utilizes the edge information to estimate the face feature positions such as eyes, nose and mouth, using a variable size face feature template, the initial size of which is predetermined by using a facial image database. The method is robust in the sense that the detection is possible with facial images with different skin color and different facial orientations. Subsequently, by using a deformable circular template matching two iris positions of the face are determined and are used in the edge pixel counting, to track the features in the next frame. Although feature tracking using gray scale template matching often fails when inter frame correlation around the feature areas are very low due to facial expression change (such as, talking, smiling, eye blinking etc.), feature tracking using edge pixel counting can track facial features reliably. Some experimental results are shown to demonstrate the effectiveness of the proposed method.

Slepian, Wolf and Wyner proved famous source coding theorems for correlated i.i.d. sources. On the other hand recently Han and Verdú have shown the source and channel coding theorems on general sources and channels whose statistics can be arbitrary, that is, no assumption such as stationarity or ergodicity is imposed. We prove source coding theorems on correlated general sources by using the method which Han and Verdú developed to prove their theorems. Also, through an example, we show some new results which are essentially different from those already obtained for the i.i.d. source cases.

The throughput performance of the non-persistent carrier sense spread spectrum with overload detection (NP-CSSS/OD) protocol is analysed and compared with that of the conventional non-persistent and one-persistent carrier sense multiple access with collision detection (NP-CSMA/CD and 1P-CSMA/CD) and the one-persistent carrier sense spread spectrum with overload detection (1P-CSSS/OD) protocols. We also introduced utilization measurements and did some performance comparisons between these protocols. At high offered loads, the NP-CSSS/OD protocol is found to offer the best throughput and utilization performances amongst them.

This paper proposes a new type of optically controlled BFN (beam forming network), an electro-optic BFN using an optical waveguide structure. In this BFN, antenna beam forming is performed using conventional optical variable phase shifters and conventional optical variable directional couplers. An electro-optic BFN can easily utilize monolithic integration capability that will be advantageous to microwave stabilization. In order to discuss practical applicability, microwave characteristics and beam forming characteristics were examined using an experimental BFN fabricated on a LiNbO3 substrate. Resulting from electro-optic lightwave control, linear phase shifting and variable amplitude distribution were measured at various microwave frequencies. Without any other control except for optical offset frequency locking and applying constant voltages, typical short term fluctuation in L-band microwave was measured to be within 3 degreesp-p in phase and 2.5 dBp-p in amplitude, respectively. For the first time, an electro-optic BFN was successful in performing beam forming in an L-band array antenna as well as coaxial cables. It was also verified that radiation pattern measured in 60 degree beam steering using the experimental BFN was comparable to that calculated using conventional microwave BFNs. The experimental results show the feasibility of utilizing an electro-optic BFN in future advanced microwave/millimeter-wave array antenna systems.

Data flow is a paradigm for concurrent computations in which a collection of parallel processes communicate asynchronously. For nondeterministic data flow networks many semantic models have been defined, however, it is complex to reason about the semantics of a network. In this paper, we introduce a transformation between data flow networks and the LOTOS specification language to make available theories and tools developed for process algebras for the semantic analysis based on traces of the networks. The transformation does not establish a one-to-one mapping between the traces of a data flow network and the LOTOS specification, but maps each network in a specification which usually contains more traces. The obtained system specification has the same set of traces as the corresponding network if they are finite, otherwise also non fair traces are included. Formal analysis and verification methods can still be applied to prove properties of the original data flow network, allowing in case of networks with finite traces to prove also network equivalence.

Distributed computing over a network of workstations continues to be an illusive goal. Its main obstacle is the delay penalty due to network protocol and OS overhead. We present in this paper a low level hardware supported scheme for managing distributed shared memory (DSM), as an underlying paradigm for distributed computing. The proposed DSM is novel in that it employs a two-tier paging scheme that reduces the probability of false sharing and facilitates an efficient hardware implementation. The scheme employs a standard OS page and divides it into fixed smaller memory units called paragraphs, similar to cache lines. This scheme manages the shared data regions only, while other regions are handled by the OS in the standard manner without modification. A hardware extension of a traditional MMU, namely Distributed MMU or DMMU, is introduced to support the DSM. Shared memory coherency is maintained through a write-invalidate protocol. An analytical model is built to evaluate the system sensitivity to various parameters and to assess its performance.