Oscar-Ortega LOBO Masayuki NUMAO
Using decision trees to fill the missing values in data has been shown experimentally to be useful in some domains. However, this is not the general case. In other domains, using decision trees for imputing missing attribute values does not outperform other methods. Trying to identify the reasons behind the success or failure of the various methods for filling missing values on different domains can be useful for deciding the technique to be used when learning concepts from a new domain with missing values. This paper presents a technique by which to approach to previous goal and presents the results of applying the technique on predicting the success or failure of a method that uses decision trees to fill the missing values in an ordered manner. Results are encouraging because the obtained decision tree is simple and it can even provide hints for further improvement on the use of decision trees to impute missing attribute values.
Tae-Young YANG Chungyong LEE Dae-Hee YOUN
A speaker adaptation technique that maximizes the observation probability of an input speech is proposed. It is applied to semi-continuous hidden Markov model (SCHMM) speech recognizers. The proposed algorithm adapts the mean µ and the covariance Σ iteratively by the gradient search technique so that the features of the adaptation speech data could achieve maximum observation probabilities. The mixture coefficients and the state transition probabilities are adapted by the model interpolation scheme. The main advantage of this scheme is that the means and the variances, which are common to all states in SCHMM, are adapted independently from the other parameters of SCHMM. It allows fast and precise adaptation especially when there is a large acoustic mismatch between the reference model and a new speaker. Also, it is possible that this scheme could be adopted to other areas which use codebook. The proposed adaptation algorithm was evaluated by a male speaker-dependent, a female speaker-dependent, and a speaker-independent recognizers. The experimental results on the isolated word recognition showed that the proposed adaptation algorithm achieved 46.03% average enhancement in the male speaker-dependent recognizer, 52.18% in the female speaker-dependent recognizer, and 9.84% in the speaker-independent recognizer.
Takuya AOKI Tatsuya MORISHITA Toshiyuki TANAKA Masao TAKI
The application of an active noise control system in a finite-length duct is studied. Previously proposed single-input-single-output systems are inappropriate in this case, because reflection at the terminals degrades the performance, and/or infinite-impulse-response filters are required for perfect noise cancellation. In this paper, we propose a single-input-single-output system applicable to finite-length ducts, which theoretically achieves perfect noise cancellation while using finite-impulse-response filters only. The tap lengths of the filters are as short as the delays between the reference sensor and the secondary source. A useful implementation of the proposed system is also discussed.
Norihiko SHINOMIYA Hiroshi TAMURA Hitoshi WATANABE
This paper deals with a study of a problem for finding the minimum-cost spanning tree with a response-time bound. The relation of cost and response-time is given as a monotonous decreasing and convex function. Regarding communication bandwidth as cost in an information network, this problem means a minimum-cost tree shaped routing for response-time constrained broadcasting, where any response-time from a root vertex to other vertex is less than a given time bound. This problem is proven to be NP-hard and consists of the minimum-cost assignment to a rooted tree and the minimum-cost tree finding. A nonlinear programming algorithm solves the former problem for the globally optimal solution. For the latter problem, different types of heuristic algorithms evaluate to find a near optimal solution experimentally.
Yoshiyuki SHINKAWA Masao J. MATSUMOTO
It is one of the difficulties in enterprise modeling that we must deal with many fragmented pieces of knowledge provided by various domain-experts, which are usually based on mutually different viewpoints of them. This paper presents a formal approach to integrate those pieces into enterprise-wide model units using Rough Set Theory (RST). We focus on business processes in order to recognize and identify the constituents or units of enterprise models, which would compose the model expressing the various aspects of the enterprise. We defined five model unit types of "resource," "organization," "task," "function," and "behavior. " The first three types represent the static aspect of the enterprise, whereas the last two types represent the dynamic aspect of it. Those units are initially elicited from each domain-expert as his/her own individual model units, then they are integrated into enterprise-wide units using RST. Our approach is methodology-free, and any methodologies can include it in their early stage to identify what composes the enterprise.
Atsushi NAKAMURA Masaki NAITO Hajime TSUKADA Rainer GRUHN Eiichiro SUMITA Hideki KASHIOKA Hideharu NAKAJIMA Tohru SHIMIZU Yoshinori SAGISAKA
This paper describes an application of a speech translation system to another task/domain in the real-world by using developmental data collected from real-world interactions. The total cost for this task-alteration was calculated to be 9 Person-Month. The newly applied system was also evaluated by using speech data collected from real-world interactions. For real-world speech having a machine-friendly speaking style, the newly applied system could recognize typical sentences with a word accuracy of 90% or better. We also found that, concerning the overall speech translation performance, the system could translate about 80% of the input Japanese speech into acceptable English sentences.
Cheng-Chung HSU Wu-Shiung FENG
This work describes a novel structural design for nth-order (n2) transconductance and grounded capacitor (Gm-C) filters with multiple loop feedback technology. The proposed design largely focuses on elucidating the relationship between the Gm-C filter structure and the feedback matrix, allowing us to systematically generate and formulate general design equations. The proposed design allows many new interesting filter configurations to be produced alongside some known structures. Furthermore, numerical design examples and simulation results verify the feasibility of the Gm-C filter approach.
Miodrag J. MIHALJEVIC Marc P. C. FOSSORIER Hideki IMAI
An algorithm for cryptanalysis of certain keystream generators is proposed. The developed algorithm has the following two advantages over other reported ones: it is more powerful, and it can be implemented by a high-speed software or a simple hardware suitable for high parallel architectures. The algorithm is based on error-correction of information bits only (of the corresponding binary block code) with a novel method for construction of the parity-checks, and the employed error-correction procedure is an APP based threshold decoding. Experimental and theoretical analyses of the algorithm performance are presented, and its complexity is evaluated. The proposed algorithm is compared with recently proposed improved fast correlation attacks based on convolutional codes and turbo decoding. The underlying principles, performance and complexity are compared, and the gain obtained with the novel approach is pointed out.
This paper proposes an algorithm that adaptively estimates time-varying noise variance used in Kalman filtering for real-time speech signal enhancement. In the speech signal contaminated by white noise, the spectral components except dominant ones in high frequency band are expected to reflect the noise energy. Our approach is first to find the dominant energy bands over speech spectrum using LPC. We then calculate the average value of the actual spectral components over the high frequency region excluding the dominant energy bands and use it as the noise variance. The resulting noise variance estimate is then applied to Kalman filtering to suppress the background noise. Experimental results indicate that the proposed approach achieves a significant improvement in terms of speech enhancement over those of the conventional Kalman filtering that uses the average noise power over silence interval only. As a refinement of our results, we employ multiple-Kalman filtering with multiple noise models and improve the intelligibility.
Hitoshi SATOH Yuji UKAI Noboru NIKI Kenji EGUCHI Kiyoshi MORI Hironobu OHMATSU Ryutarou KAKINUMA Masahiro KANEKO Noriyuki MORIYAMA
In this paper, we present a computer-aided diagnosis (CAD) system to automatically detect lung cancer candidates at an early stage using a present and a past helical CT screening. We have developed a slice matching algorithm that can automatically match the slice images of a past CT scan to those of a present CT scan in order to detect changes in the lung fields over time. The slice matching algorithm consists of two main process: the process of extraction of the lungs, heart, and descending aorta and the process of matching slices of the present and past CT images using the information of the lungs, heart, and descending aorta. To evaluate the performance of this algorithm, we applied it to 50 subjects (total of 150 scans) screened between 1993 and 1998. From these scans, we selected 100 pairs for evaluation (each pair consisted of scans for the same subject). The algorithm correctly matched 88 out of the 100 pairs. The slice images for the present and past CT scans are displayed in parallel on the CRT monitor. Feature measurements of the suspicious regions are shown on the relevant images to facilitate identification of changes in size, shape, and intensity. The experimental results indicate that the CAD system can be effectively used in clinical practice to increase the speed and accuracy of routine diagnosis.
Shinichi YOROZU Yoshio KAMEDA Shuichi TAHARA
Within the next few decades, high-end telecommunication systems on the larger nationwide network will require a switching capacity of over 5 Tbps. Advanced optical transmission technologies, such as wavelength division multiplexing (WDM) will support optical-fiber data transmission at such speeds. However, semiconductors may not be capable of high-throughput data switching because of the limitations by power consumption and operating speed, and pin count. Superconducting single flux quantum (SFQ) technology is a promising approach for overcoming these problems. This paper proposed an optical-electrical-SFQ hybrid switching system and a novel switch architecture. This architecture uses time-shifted internal speedup, shuffle and grouping exchange and a Batcher-Banyan switch. Our proposed switch consists of an interface circuit with small buffers, a Batcher sorter, a time-shift-speedup buffer (TSSB), a Banyan switch, and a slowdown buffer. Simulations showed good scalability up to 100 Tbps, which no router could ever offer such features.
In this paper, we consider the complexity of recognizing ordered tree-shellable Boolean functions when Boolean functions are given as OBDDs. An ordered tree-shellable function is a positive Boolean function such that the number of prime implicants equals the number of paths from the root node to a 1-node in its ordered binary decision tree representation. We show that given an OBDD, it is possible to check within polynomial time if the function is ordered tree-shellable with respect to the variable ordering of the OBDD.
Yoshito SHUTO Hirotsugu SATO Shuichi YANAGI Masayoshi OHNO Shin SUMIDA Shunichi TOHNO
We examined the creep properties and hazard rates of plastic split alignment sleeves to ensure the long-term reliablity of optical fiber connections. It required a gauge retention force Fr of more than 200 gf to suppress the fluctuation in the insertion loss of a plastic sleeve. From the fluctuation data, we estimated the time-to-failure tf at which the Fr value became 200 gf. We estimated the acceleration parameters, median lifetimes ξ, and hazard rates λ by using the tf values based on the Weibull statistics. The ξ values decreased rapidly with increasing temperature and relative humidity. Small λ values of < 0.01 FITs and of 1 FITs were expected for 20 years in a normal atmosphere (25C/50%RH) and in a more severe case of 25C/90%RH or 45C/50%RH.
Naoto SONE Masami MOHRI Masakatu MORII Hiroshi SASANO
New good convolutional codes with optimal free distance are tabulated for the number of memories M 22 and rate R=1/2, which were selected based on the criterion of minimizing the decoding error rate and bit error rate. Furthermore, for R=1/3, 1/4 and M 13, we give the new good codes and make clear the existance of the codes with minimum free distance which achieve to Heller's upper bound for M 16.
Satoshi UNO Hideki TODE Koso MURAKAMI
B-ISDN is expected to be applied in the near future to video delivery systems for the broadcast of news and television programs. The demand for such services is increasing, and in particular, on-demand services are becoming more desirable. On-demand services allow viewers to request their favorite programs at the time that is convenient, hence catering for the wide range of modern lifestyles. As for on-demand services, there already exist Video on Demand (VoD) systems such as the original VoD or Near VoD. However, such systems have not yet been widely implemented because of the inefficient cost of communication resources, and storage. The authors' research is aimed at producing an efficient VoD system based on a high speed network. We are focused in particular on the forms of data transmission, and in this paper, we propose a new VoD system called Burst VoD. Burst VoD aggressively utilizes the multicasting technique, and involves dividing the program resource data into block files and transmitting them to viewer terminals as burst traffic over a high speed network. Simulation results comparing Burst VoD with conventional VoD show that Burst VoD achieves lower request blocking rates, efficient utilization of networks with multicasting, and almost on-demand response time to requests.
This paper deals with deadlock and fairness issues that may arise when network users request resources for guaranteed service with the resource reservation protocol (RSVP). A deadlock occurs when a request can only be satisfied if the resources reserved for another request are released, but the reserved resources are never released. The fairness issue occurs when some reservation requests may be satisfied but only after a very long wait. Our approach to these issues is based on our belief that a network should provide stable throughput and fairness whatever the behavior of the user. Our methods are unique in two respects. First, during the session setup phase, a node directly connected to the requesting users terminates the users' behavior and makes reservations fairly and efficiently in place of the users. Second, our three admission control methods allocate resources for each reservation request by considering not only the current residual bandwidth but also the properties of the requesting session; e.g., its weight (the number of resources it requires) or its age (how long it has been waiting for session setup). Our methods do not maximize the throughput since they always keep a certain amount of resources unreserved for fairness. From simulation results, however, they do provide quite fair behavior, and their throughput is stable regardless of the network size and the session holding time.
This letter points out some flaws in the previous works on UKS (unknown key-share) attacks. We show that Blake-Wilson and Menezes' revised STS-MAC (Station-to-Station Message Authentication Code) protocol, which was proposed to prevent UKS attack, is still vulnerable to a new UKS attack. Also, Hirose and Yoshida's key agreement protocol presented at PKC'98 is shown to be insecure against public key substitution UKS attacks. Finally, we discuss countermeasures for such UKS attacks.
High-speed systems require a wide-frequency-range clock system for data processing. Phase-locked loop (PLL) is used for such a system that requires wide-range variable frequency clock. Frequency calibration method enables the voltage-controlled oscillator (VCO) in a PLL to cover the expected frequency range for high-speed applications that require a wide locking range. Frequency range adjustment is implemented by means of a current digital to analog converter (DAC), which controls the performance curves of a VCO and a bias circuit. This method adjusts the VCO's frequency-voltage performance curves before functional operation so that a PLL can cover requested frequency range with its best condition. Both the limit of control voltage and its target reference voltage are given with same voltage reference. This ensures correct performance after frequency adjustment even under the temperature fluctuation. It eliminates post-production physical adjustment such as fuse trimming which increases the cost and TAT in manufacturing and testing. A high-speed wide-locking range VCO with an automatic frequency performance calibration circuit is implemented within small space in a high-speed hard disk drive channel with 0.25-µm 2.5 V CMOS four-layer metal technology.
An automated method for cryptanalysis of DFT-based analog speech scramblers is presented through statistical estimation treatments. In the proposed system, the ciphertext only attack is formulated as a combinatorial optimization problem leading to a search for the most likely key estimate. For greater efficiency, we also explore the benefits of genetic algorithm to develop an estimation method which takes into account the doubly stochastic characteristics of the underlying keyspace. Simulation results indicate that the global explorative properties of genetic algorithms make them very effective at estimating the most likely permutation and by using this estimate significant amount of the intelligibility can be recovered from the ciphertext following the attack on DFT-based speech scramblers.
A 3.3 V CMOS PLL (Phase Locked loop) with a self-feedback VCO (Voltage Controlled Oscillator) is designed for a high frequency, low voltage, and low power applications. This paper proposes a new PLL architecture to improve voltage-to-frequency linearity of VCO with a new delay cell. The proposed VCO with a self-feedback path operates at a wide frequency range of 30 MHz-1 GHz with a good linearity. The DC-DC Voltage Up/Down Converter is newly designed to regulate the control voltage of the two-stage VCO. The designed PLL architecture is implemented on a 0.6 µm n-well CMOS process. The simulation results illustrate a locking time of 2.6 µsec at 1 GHz, lock in range of 100 MHz-1 GHz, and a power dissipation of 112 mW.