In the last three decades of the 20th Century, research in speech recognition has been intensively carried out worldwide, spurred on by advances in signal processing, algorithms, architectures, and hardware. Recognition systems have been developed for a wide variety of applications, ranging from small vocabulary keyword recognition over dial-up telephone lines, to medium size vocabulary voice interactive command and control systems for business automation, to large vocabulary speech dictation, spontaneous speech understanding, and limited-domain speech translation. Although we have witnessed many new technological promises, we have also encountered a number of practical limitations that hinder a widespread deployment of applications and services. On one hand, fast progress was observed in statistical speech and language modeling. On the other hand only spotty successes have been reported in applying knowledge sources in acoustics, speech and language science to improving speech recognition performance and robustness to adverse conditions. In this paper we review some key advances in several areas of speech recognition. A bottom-up detection framework is also proposed to facilitate worldwide research collaboration for incorporating technology advances in both statistical modeling and knowledge integration into going beyond the current speech recognition limitations and benefiting the society in the 21st century.

In this paper, we propose a floating resistor circuit with positive and negative resistance operating at the low supply voltages 1.5 V. Only two transistors are connected between supply lines in order to operate under the low power supply voltages. In this circuit, current subtraction is carried out at the gate terminal for which input/output voltage is applied. As a result, the proposed circuit can realize the large range of resistance of positive and negative resistances. Therefore, in an application, the proposed circuit is used in neuro-based limit cycle generator as synaptic weights.

This paper describes truncated and impossible differentials of Feistel block ciphers with round functions of 2-layer SPN (Substitution and Permutation Network) transformation modules such as the 128-bit block cipher Camellia, which was proposed by NTT and Mitsubishi Electric Corporation. Our work improves on the best known truncated and impossible differentials, and has found a nontrivial 9-round truncated differential that may lead to a possible attack against a reduced-round version of Camellia without input/output whitening, FL or FL-1 (Camellia-NFL), in the chosen plain text scenario. Previously, only 6-round differentials were known that may suggest a possible attack of Camellia-NFL reduced to 8-rounds. We also show a nontrivial 7-round impossible differential, whereas only a 5-round impossible differential was previously known. We also consider the truncated differential of a reduced-round version of Camellia (Camellia-DS) whose round functions are composed of D-S (Diffusion and Substitution) transformation modules and without input/output whitening, FL or FL-1 (Camellia-DS-NFL), and show a nontrivial 9-round truncated differential, which may lead to a possible attack in the chosen plain text scenario. This truncated differential is effective for general Feistel structures with round functions composed of S-D (Substitution and Diffusion) or D-S transformation.

Face recognition provides an important means for realizing a man-machine interface and security. This paper presents "Smartface," a PC-based face recognition system using a temporal image sequence. The face recognition engine of the system employs a robust facial parts detection method and a pattern recognition algorithm which is stable against variations of facial pose and expression. The functions of Smartface include (i) screensaver with face recognition, (ii) customization of PC environment, and (iii) real-time disguising, an entertainment application. The system is operable on a portable PC with a camera and is implemented only with software; no image processing hardware is required.

OPKFDDs (Ordered Pseudo-Kronecker Functional Decision Diagrams) are a data structure that provides compact representation of Boolean functions. The size of OPKFDDs depends on a variable ordering and on decomposition type choices. Finding an optimal representation is very hard and the size of the search space is n! 32n-1, where n is the number of input variables. To overcome the huge search space of the problem, a genetic algorithm is proposed for the generation of OPKFDDs with minimal number of nodes.

This paper describes special-purpose hardware for large-scale logic simulation, called SP2, which executes an event driven algorithm and can simulate up to sixteen million gates. SP2 was developed, in 1992, for system verification of large-scale computer designs as a successor to SP1, which was developed in 1987. SP2 provides enhanced performance, throughput, and delay accuracy over SP1. Since 1992, SP2 has been widely used for system-level simulation of mainframes, super computers, UNIX servers and microprocessors. It is used as a powerful simulator, in all stages of design verification, or in early stages, before regression testing, by using emulators.

A shared binary decision diagram (SBDD) represents a multiple-output function, where nodes are shared among BDDs representing the various outputs. A partitioned SBDD consists of two or more SBDDs that share nodes. The separate SBDDs are optimized independently, often resulting in a reduction in the number of nodes over a single SBDD. We show a method for partitioning a single SBDD into two parts that reduces the node count. Among the benchmark functions tested, a node reduction of up to 23% is realized.

A postprocessing algorithm is presented for blocking artifact reduction in block-coded images using the adaptive filters along the pattern of neighborhood blocks. Blocking artifacts appear as irregular high-frequency components at block boundaries, thereby reducing the noncorrelation between blocks due to the independent quantization process of each block. Accordingly, block-adaptive filtering is proposed to remove such components and enable similar frequency distributions within two neighborhood blocks and a high correlation between blocks. This type of filtering consists of inter-block filtering to remove blocking artifacts at the block boundaries and intra-block filtering to remove ringing noises within a block. First, each block is classified into one of seven classes based on the characteristics of the DCT coefficient and MV (motion vector) received in the decoder. Thereafter, adaptive intra-block filters, approximated to the normalized frequency distributions of each class, are applied adaptively according to the various patterns and frequency distributions of each block as well as the filtering directions in order to reduce the blocking artifacts. Finally, intra-block filtering is performed on those blocks classified as complex to reduce any ringing noise without blurring the edges. Experimental tests confirmed the effectiveness of the proposed algorithm.

The optimization of path bandwidth allocation in large-scale telecommunication networks is studied. By introducing a decomposition-coordination scheme to global optimization of the path bandwidth allocation which aims at minimizing the worst case call blocking probabilities in the network, the spatial and time complexities are both reduced, while the accuracy is almost the same as that given by direct optimization.

Functional decomposition is an essential technique of logic synthesis and is important especially for FPGA design. Bertacco and Damiani proposed an efficient algorithm finding simple disjoint decomposition using Binary Decision Diagrams (BDDs). However, their algorithm is not complete and does not find all the decompositions. This paper presents a complete theory of simple disjoint decomposition and describes an efficient algorithm using BDDs.

The one-sample locally optimum rank detector test statistics for composite signals in multiplicative and signal-dependent noise are obtained. Since the one-sample locally optimum rank detector makes use of the sign statistics of observations as well as the rank statistics, both 'even' and 'odd' score functions have to be considered. Although the one-sample locally optimum rank detector requires two score functions while the two-sample detector requires only one score function, the one-sample detector requires fewer calculations since it has to rank fewer observations.

This paper presents a method of extracting car license plates captured from the arbitrary directions by using symmetry features. The generalized symmetry transform (GST) produces continuous features of symmetry between two points by combining locality constraint and reflectional symmetry, but its time complexity of GST is increased by the second power of the radius of a searching window. To achieve considerable reduction of this time complexity, we propose a scan line based GST (SLGST) which calculates the symmetry between a pair of edge pixels along the scan lines. Instead of calculating the global symmetry of a license plate, we assign the symmetry contribution acquired from SLGST to the corner point estimated from two edge points and their gradient orientations. The right angle tuned SLGST (RATSLGST) is newly proposed to detect the right angle corners more effectively. Image normalization by image warping is adopted to make such segmentation of license plate and later identification much easier. We also adopt the verifier that evaluates a candidate license plate to enhance extraction rate. Our experiments show that the proposed method for extracting perspectively distorted license plates is fairly reliable.

This paper discusses encoding of vehicular position information using predictive algorithms in inter-vehicle communications (IVC) from the viewpoints of source coding and noisy channels. Two vehicular driving models are assumed; one is the 15-mode as a suburban rapid transit driving pattern, the other is called calming mode as a street-driving pattern. Three types of schemes are compared; a pulse code modulation (PCM) scheme, a predictive coding (PC) scheme, and the variable interval prediction (VIP) scheme that is proposed here. This paper assumes that precise position information is got from a positioning system, and that all the transmitters and receivers have common predictors. Performance comparisons of the three types of schemes are carried out both of noiseless and noisy channels. Results show that the VIP scheme is superior to any other scheme.

In this paper, we consider a flexible method for designing n-entities communication protocols and services. The proposed technique considers alternative and parallel composition of n service specifications and n protocol specifications, where n 2. The specifications are specified in Basic LOTOS which is a Formal Description Technique (FDT). We use the weak bisimulation equivalence () to represent the correctness properties between the service specification and the protocol specification.

DS/CDMA systems employing long-period PN sequences are becoming a widespread standard of wireless communication systems. However, fast acquisition of long-period PN sequences at a low hardware cost is conventionally a difficult problem. This paper proposes a new fast acquisition algorithm for a class of PN sequences, which includes m- and GMW sequences as special cases, and shows that the mean (correct) acquisition time can be considerably reduced under input SNR values well below those used in modern DS/CDMA systems. Its fast acquisition capability is based on a decomposition of long PN sequences into a number of short ones and achieves a significantly reduced code phase uncertainty of acquisition at relatively small hardware cost. It can be applied as a (part of) acquisition system of a DS/CDMA system instead of a slow sliding correlator or a costly matched filter.

The two-sample locally optimum rank detector test statistics for composite signals in additive, multiplicative, and signal-dependent noise are obtained in this letter. Compared with the structure of the one-sample locally optimum rank detector, that of the two-sample locally optimum rank detector is shown to be simpler, although it needs more computations. It is known that there is a trade-off of computational complexity and structural simplicity between the one- and two-sample detectors.

The authors propose a novel 3-way power divider named a planar cross-junction, which is used as the center feed for single-layer slotted waveguide arrays. A feeding waveguide consisting of a cascade connection of these dividers is placed at the middle of radiating waveguide in a single layer. The length of radiating waveguides is halved; the long line effect in traveling wave operation is halved and bandwidth is widened. One divider as a unit is designed by Galerkin's method of moments to suppress the reflection and to control the amplitude and the phase of the divided power into two radiating waveguides on both sides of a feed one. Two types of the cross-junction with a different divided power ratio are designed and tested by experiments in 4 GHz band. The mutual coupling effects between two adjacent cross-junctions as cascaded in a feeding waveguide of the array are predicted to be small enough; units designed here are directly applicable for a multiple-way power divider.

Co-containing ferrite thin-film media deposited by a reactive-ECR-sputtering at a low substrate temperature of 150 degree Celsius were oxidized by ECR plasma. The magnetic properties and recording characteristics of the media were improved by the oxidation with maintaining a smooth surface. The media showed high D50 of 203 kFRPI in MIG head recording and reproduction. The Co-containing ferrite thin-film is feasible to be used as a protective overcoat layer.

The lossy magnetic composite material made from soft magnetic metal powder and rubber is widely used as an EMI countermeasure material, due to its higher magnetic loss than those of spinel ferrites in microwave frequencies. In this paper, we clarify the material characteristics by measuring the relative complex permeability r and relative complex permittivity r of two kinds of composite materials in microwave frequencies. Since the composite materials are anisotropic, both r and r are measured as diagonal tensors by utilizing extended S-parameter method. The results show that the imaginary part of r of flaky-powder composite exceeded the Snoek's limit for the spinel ferrites which has been reported so far. The measured r and r are partially compared with those measured by cavity resonator method, and good agreement is obtained.

We propose a high-sensitivity and wide-dynamic-range position sensor using logarithmic-response and correlation circuit. The 3-D measurement system using the proposed position sensor has advantages to applications, for example a walking robot and a recognition system on vehicles, which require both of availability in various backgrounds and safe light projection for human eyes. The position sensor with a 64 64 pixel array has been developed and successfully tested. We describe the sensitivity of position detection as SBR (Signal-to-Background Ratio). The minimum SBR of the sensor is -13.9 dB lower than standard sensors. High sensitivity under -10 dB SBR is realized in a dynamic range of 41.7 dB in terms of background illumination. Experimental results of position detection and 3-D measurement in a strong background illumination are also presented.