This paper describes a novel parameter generation algorithm for an HMM-based speech synthesis technique. The conventional algorithm generates a parameter trajectory of static features that maximizes the likelihood of a given HMM for the parameter sequence consisting of the static and dynamic features under an explicit constraint between those two features. The generated trajectory is often excessively smoothed due to the statistical processing. Using the over-smoothed speech parameters usually causes muffled sounds. In order to alleviate the over-smoothing effect, we propose a generation algorithm considering not only the HMM likelihood maximized in the conventional algorithm but also a likelihood for a global variance (GV) of the generated trajectory. The latter likelihood works as a penalty for the over-smoothing, i.e., a reduction of the GV of the generated trajectory. The result of a perceptual evaluation demonstrates that the proposed algorithm causes considerably large improvements in the naturalness of synthetic speech.

A numerical hybrid method for analyzing the wireless channel of Multiple-Input Multiple-Output (MIMO) communication system is proposed by combining of the method of moments (MoM) and the finite difference time domain (FDTD) method. The proposed method is capable of investigating a more practical MIMO wireless channel than the conventional methods, and CPU time is much less than that of the FDTD method in analysis of spatial statistical characteristics of received signals. Based on the channel transfer matrix obtained by the proposed method, the wall effect on indoor MIMO channel capacity are investigated with consideration of received power, Ricean K-factor and effective degrees of freedom (EDOF) of multipaths by changing the wall locations and material.

The physical origin of stress-induced leakage currents (SILC) in ultra-thin SiO2 films is described. Assuming a two-step trap-assisted tunneling process accompanied with an energy relaxation process of trapped electrons, conditions of trap sites which are origin of SICL are quantitatively found. It is proposed that the trap site location and the trap state energy can be explained by a mean-free-path of hole in SiO2 films and an atomic structure of the trap site by the O vacancy model.

In this paper we give a simple algorithm to generate all partitions of a positive integer n. The problem is one of the basic problems in combinatorics, and has been extensively studied for a long time. Our algorithm generates each partition of a given integer in constant time for each without repetition, while best known algorithm generates each partition in constant time on "average." Also, we propose some algorithms to generate all partitions of an integer with some additional property in constant time.

Monolithically integrated four-channel distributed feedback (DFB) laser array has been fabricated by metal organic vapor phase epitaxy (MOVPE) selective area growth for 1.55 µm coarse-wavelength division multiplexing (CWDM) systems. Wide-stripe MOVPE selective area growth and electron-beam lithography are used to obtain wide CWDM channel spacing of 20 nm. Compared to hybrid integration of discrete lasers, monolithic integration of laser array on a single substrate greatly simplifies device alignment and packaging process.

This paper proposes a simple and feasible decision-feedback channel tracking scheme for multiple-input multiple-output orthogonal frequency division multiplexing (MIMO-OFDM) systems designed for wireless local area networks (LANs). In the proposed scheme, the channel state matrix for each subcarrier is tentatively estimated from a replica matrix of the transmitted signals. The estimated channel matrices, each derived at a different timing, are combined, and the previously estimated channel matrices are replaced with the latest ones. Unlike conventional channel tracking schemes based on a Kalman filter, the proposed scheme needs no statistical information about a MIMO channel, which makes the receiver structure quite simple. The packet error rate (PER) performances for the proposed scheme are evaluated on computer simulations. When there are three transmit and receive antennas, the subcarrier modulation scheme is 64 QAM, and the coding rate is 3/4, the proposed scheme keeps the SNR degradation at PER of 1e-2 less than 0.1 dB when the velocity of receiver is 3 km/h in an indoor office environment at 5 GHz band. In addition, compared to the conventional channel tracking scheme based on known pilot symbols, the proposed scheme improves throughput performance by 13.8% because it does not need pilot symbols. These results demonstrate that the proposed channel tracking scheme is simple and feasible for implementation in MIMO-OFDM systems based on wireless LANs.

We propose a novel paging scheme with a variable paging interval for low power consumption and/or short paging delay. The proposed scheme is based on the fact that packet arrivals during a session follow the characteristics of self-similar process for Http service, while session arrival statistics can be modeled as the Poisson process. The adjustment of paging period provides a useful solution for efficient paging to the UE in the dormant state on packet-switched cellular networks, even though the paging performance is strongly dependent on the traffic arrival model.

This paper presents a design method for AND-OR-EXOR three-level networks, where a single two-input exclusive-OR (EXOR) gate is used. The network realizes an EXOR of two sum-of-products expressions (EX-SOPs). The problem is to minimize the total number of products in the two sum-of-products expressions (SOPs). We introduce the notion of µ-equivalence of logic functions to develop exact minimization algorithms for EX-SOPs with up to five variables. We minimized all the NP-representative functions for up to five variables and showed that five-variable functions require 9 or fewer products in minimum EX-SOPs. For n-variable functions, minimum EX-SOPs require at most 9·2n-5 (n ≤ 6) products. This upper bound is smaller than 2n-1, which is the upper bound for SOPs. We also found that, for five-variable functions, on the average, minimum EX-SOPs require about 40% fewer literals than minimum SOPs.

In this paper, we propose a new preemptive scheme with release message in optical burst switching (OBS) networks. In the proposed scheme, when a low priority burst is preempted at some intermediate node, two RELEASE messages are sent immediately from the intermediate node to both source and destination nodes (two-way release message transmission), and the RELEASE messages release the corresponding wavelengths for the preempted burst. We consider six wavelength selection rules for the preemption and evaluate the performances of the selection rules by simulations. Numerical examples show that our scheme utilizes wavelengths effectively and, with the optimal selection rule, can decrease the burst loss probability in a large-scale DWDM network.

In this paper, we propose a useful algorithm that can be applied to reduce the response time of speech recognizers based on HMM's. In our algorithm, to reduce the response time, promising HMM states are selected by single Gaussians. In speech recognition, HMM state likelihoods are evaluated by the corresponding single Gaussians first, and then likelihoods by original full Gaussians are computed and replaced only for the HMM states having relatively large likelihoods. By doing so, we can reduce the pattern-matching time for speech recognition significantly without any noticeable loss of the recognition rate. In addition, we cluster the single Gaussians into groups by measuring the distance between Gaussians. Therefore, we can reduce the extra memory much more. In our 10,000 word Korean POI (point-of-interest) recognition task, our proposed algorithm shows 35.57% reduction of the response time in comparison with that of the baseline system at the cost of 10% degradation of the WER.

A statistical speech synthesis system based on the hidden Markov model (HMM) was recently proposed. In this system, spectrum, excitation, and duration of speech are modeled simultaneously by context-dependent HMMs, and speech parameter vector sequences are generated from the HMMs themselves. This system defines a speech synthesis problem in a generative model framework and solves it based on the maximum likelihood (ML) criterion. However, there is an inconsistency: although state duration probability density functions (PDFs) are explicitly used in the synthesis part of the system, they have not been incorporated into its training part. This inconsistency can make the synthesized speech sound less natural. In this paper, we propose a statistical speech synthesis system based on a hidden semi-Markov model (HSMM), which can be viewed as an HMM with explicit state duration PDFs. The use of HSMMs can solve the above inconsistency because we can incorporate the state duration PDFs explicitly into both the synthesis and the training parts of the system. Subjective listening test results show that use of HSMMs improves the reported naturalness of synthesized speech.

We consider the capacitated multi-source multicast tree routing problem (CMMTR) in an undirected graph G=(V,E) with a vertex set V, an edge set E and an edge weight w(e) ≥ 0, e ∈ E. We are given a source set S ⊆ V with a weight g(e) ≥ 0, e ∈ S, a terminal set M ⊆ V-S with a demand function q : M → R+, and a real number κ > 0, where g(s) means the cost for opening a vertex s ∈ S as a source in a multicast tree. Then the CMMTR asks to find a subset S′⊆ S, a partition {Z1,Z2,...,Zl} of M, and a set of subtrees T1,T2,...,Tl of G such that, for each i, ∑t∈Ziq(t) ≤ κ and Ti spans Zi∪{s} for some s ∈ S′. The objective is to minimize the sum of the opening cost of S′and the constructing cost of {Ti}, i.e., ∑s∈S′g(s)+w(Ti), where w(Ti) denotes the sum of weights of all edges in Ti. In this paper, we propose a (2ρUFL+ρST)-approximation algorithm to the CMMTR, where ρUFL and ρST are any approximation ratios achievable for the uncapacitated facility location and the Steiner tree problems, respectively. When all terminals have unit demands, we give a ((3/2)ρUFL+(4/3)ρST)-approximation algorithm.

One approach to accommodating IP traffic on a wavelength division multiplexing (WDM) network is to construct a logical topology, establishing a set of lightpaths between nodes. The lightpaths carry IP traffic but do not require any electronic packet processing at intermediate nodes, thereby reducing the load on those nodes. When the IP and WDM networks have independent routing functions, however, the lightpaths in the WDM network may not be fully utilized by the IP router. It is therefore necessary to integrate the two routing mechanisms in order to utilize resources efficiently and adapt to changes in traffic. In this paper, we propose an integrated routing mechanism for IP over WDM networks. The key idea is to first prepare a set of virtual-links representing the lightpaths that can be established by the WDM network, then calculate the minimum cost route on an IP network including those links. Our simulation results show that when traffic patterns do not change, the throughput of our method is almost the same as that of a logical topology optimally designed for a given traffic demand. When traffic patterns change, the throughput of our method is about 50% higher than that of the logical topology.

In recent years, the inherent effectiveness of Peer-to-Peer (P2P) networks has been advocated to address scalability issues in large scale Internet-based on-Demand streaming services. Most of existing works adopt Cache-and-Relay (CR) scheme to exploit a cooperative paradigm among peers. In this paper, we mainly present our practical evaluation study of the scalability of the CR scheme by taking into account of more than 20,000,000 collected real traces. Based on trace-driven simulations, we conclude that the CR scheme is not as effective as previously reported in terms of saving server bandwidth.

The model selection for neural networks is an essential procedure to get not only high levels of generalization but also a compact data model. Especially in terms of getting the compact model, neural networks usually outperform other kinds of machine learning methods. Generally, models are selected by trial and error testing using whole learning samples given in advance. In many cases, however, it is difficult and time consuming to prepare whole learning samples in advance. To overcome these inconveniences, we propose a hybrid on-line learning system for a radial basis function (RBF) network that repeats quick learning of novel instances by rote during on-line periods (awake phases) and repeats pseudo rehearsal for model selection during out-of-service periods (sleep phases). We call this system Incremental Learning with Sleep (ILS). During sleep phases, the system basically stops the learning of novel instances, and during awake phases, the system responds quickly. We also extended the system so as to shorten the periodic sleep periods. Experimental results showed the system selects more compact data models than those selected by other machine learning systems.

In this letter, we propose a new modification to the belief propagation (BP) decoding algorithm for Finite-Geometry low-density parity-check (LDPC) codes. The modification is based on introducing feedback into the iterative process, which can break the oscillations of bit log-likelihood ratio (LLR) values. Simulations show that, with a given maximum iteration, the "feedback BP" (FBP) algorithm can achieve better performance than the conventional belief propagation algorithm.

Silva and Krogh formulate a sampled-data hybrid automaton to deal with time-driven events and discuss its verification. In this paper, we consider a state feedback control problem of the automaton. First, we introduce two transition systems as semantics of the automaton. Next, using these transition systems, we derive necessary and sufficient conditions for a predicate to be control-invariant. Finally, we show that there always exists the supremal control-invariant subpredicate for any predicate.

For more efficient data transmissions, a new MLP/BP-based channel equalizer is proposed to compensate for multi-path fading in wireless applications. In this work, for better system performance, we apply the soft output and the soft feedback structure as well as the soft decision channel decoding. Moreover, to improve packet error rate (PER) and bit error rate (BER), we search for the optimal scaling factor of the transfer function in the output layer of the MLP/BP neural networks and add small random disturbances to the training data. As compared with the conventional MLP/BP-based DFEs and the soft output MLP/BP-based DFEs, the proposed MLP/BP-based soft DFEs under multi-path fading channels can improve over 3-0.6 dB at PER=10-1 and over 3.3-0.8 dB at BER=10-3.

Some major P2P file distribution systems adopt Tit-For-Tat exchange strategy, which means "initially cooperate, then respond in kind to a previous opponent's action, i.e. cooperative or not." However, when sharing a file on such P2P systems, the random peer selection has a problem in that each peer cannot download the file enough efficiently. The peer selection method that groups peers according to their rate has been proposed to solve this problem. This method is supposed to be able to alleviate the difference in performance among peers because it lets peers with similar transmitting rate connect to each other. However, when reduction in peer performance or link one occurs, which is often observed on today's Internet, some problems will emerge, such as it takes a long time for the existing method to reconfigure groups; 2) immediate reconstruction of neighbor peers has not been taken into account when peers detect deterioration in downloading performance. Therefore, we propose a method that reconfigures the group of neighbor peers once a peer notices that the performance of connected peers decreases. The proposed method is evaluated through simulation experiments using BitTorrent as an instance of Tit-For-Tat strategy. The download time of all peers and that of the peer with performance deterioration are estimated focusing on the effect of switching a degraded peer to another immediately. As a result, we confirm that our proposal can distribute files among all peers faster than the existing method keeping incentives for users to some extent. We believe that the proposal which can adapt to the sudden network deterioration is one of the most important technologies for evolution of network software.

In this paper we focus on building a large scale keyword search service over structured Peer-to-Peer (P2P) networks. Current state-of-the-art keyword search approaches for structured P2P systems are based on inverted list intersection. However, the biggest challenge in those approaches is that when the indices are distributed over peers, a simple query may cause a large amount of data to be transmitted over the network. We propose in this paper a new P2P keyword search scheme, called "Proof," which aims to reduce the network traffic generated during the intersection process. We applied three main ideas in Proof to reduce network traffic, including (1) using a sorted query flow, (2) storing content summaries in the inverted lists, and (3) setting a stop condition for the checking of content summaries. We also discuss the advantages and limitations of Proof, and conducted extensive experiments to evaluate the search performance and the quality of search results. Our simulation results showed that, compared with previous solutions, Proof can dramatically reduce network traffic while providing 100% precision and high recall of search results, at some additional storage overhead.