In this paper, the simplified search designs for the stochastic codebook of algebraic code excited linear prediction (ACELP) for ITU-T G.729D speech coder are proposed. By using two search rounds and limiting the search range, the computational complexity of the proposed approach is only 6.25% of the full search method recommended by G.729D. In addition, the computational complexity of proposed approach is only 59% of the global pulse replacement search method recommended by G.729.1. Simulation results show that the coded speech quality evaluated by using the standard subjective and objective quality measurements is with perceptually negligible degradation.

In this paper we show some new look at large deviation theorems from the viewpoint of the information-spectrum (IS) methods, which has been first exploited in information theory, and also demonstrate a new basic formula for the large deviation rate function in general, which is expressed as a pair of the lower and upper IS rate functions. In particular, we are interested in establishing the general large deviation rate functions that are derivable as the Fenchel-Legendre transform of the cumulant generating function. The final goal is to show, under some mild condition, a necessary and sufficient condition for the IS rate function to be derivable as the Fenchel-Legendre transform of the cumulant generating function, i.e., to be a rate function of Gartner-Ellis type.

In this paper we present our investigation into improving the performance of our computer-assisted language learning (CALL) system through exploiting the acoustic model and features within the speech recognition framework. First, to alleviate channel distortion, speaker-dependent cepstrum mean normalization (CMN) is adopted and the average correlation coefficient (average CC) between machine and expert scores is improved from 78.00% to 84.14%. Second, heteroscedastic linear discriminant analysis (HLDA) is adopted to enhance the discriminability of the acoustic model, which successfully increases the average CC from 84.14% to 84.62%. Additionally, HLDA causes the scoring accuracy to be more stable at various pronunciation proficiency levels, and thus leads to an increase in the speaker correct-rank rate from 85.59% to 90.99%. Finally, we use maximum a posteriori (MAP) estimation to tune the acoustic model to fit strongly accented test speech. As a result, the average CC is improved from 84.62% to 86.57%. These three novel techniques improve the accuracy of evaluating pronunciation quality.

In this paper, a new lifting-based shape-direction-adaptive discrete wavelet transform (SDA-DWT) which can be used for arbitrarily shaped segments is proposed. The SDA-DWT contains three major techniques: the lifting-based DWT, the adaptive directional technique, and the concept of object-based compression in MPEG-4. With SDA-DWT, the number of transformed coefficients is equal to the number of pixels in the arbitrarily shaped segment image, and the spatial correlation across subbands is well preserved. SDA-DWT also can locally adapt its filtering directions according to the texture orientations to improve energy compaction for images containing non-horizontal or non-vertical edge textures. SDA-DWT can be applied to any application that is wavelet based and the lifting technique provides much flexibility for hardware implementation. Experimental results show that, for still object images with rich orientation textures, SDA-DWT outperforms SA-DWT up to 5.88 dB in PSNR under 2.15-bpp (bit / object pixel) condition, and reduces the bit-budget up to 28.5% for lossless compression. SDA-DWT also outperforms DA-DWT up to 5.44 dB in PSNR under 3.28-bpp condition, and reduces the bit-budget up to 14.0%.

This paper presents a text-independent speaker verification method using Gaussian mixture models (GMMs), where only utterances of enrolled speakers are required. Artificial cohorts are used instead of those from speaker databases, and GMMs for artificial cohorts are generated by changing model parameters of the GMM for a claimed speaker. Equal error rates by the proposed method are about 60% less than those by a conventional method which also uses only utterances of enrolled speakers.

We have previously proposed a diffusion-type flow control mechanism as a solution for severely time-sensitive flow control required for high-speed networks. In this mechanism, each node in a network manages its local traffic flow using the basis of only the local information directly available to it, by using predetermined rules. In addition, the implementation of decision-making at each node can lead to optimal performance for the whole network. Our previous studies show that our flow control mechanism with certain parameter settings works well in high-speed networks. However, to apply this mechanism to actual networks, it is necessary to clarify how to design a parameter in our control mechanism. In this paper, we investigate the range of the parameter and derive its optimal value enabling the diffusion-type flow control to work effectively.

Spectral subtraction is commonly used for speech enhancement in a single channel system because of the simplicity of its implementation. However, this algorithm introduces perceptually musical noise while suppressing the background noise. We propose a wavelet-based approach in this paper for suppressing the background noise for speech enhancement in a single channel system. The wavelet packet transform, which emulates the human auditory system, is used to decompose the noisy signal into critical bands. Wavelet thresholding is then temporally adjusted with the noise power by time-adapted noise estimation. The proposed algorithm can efficiently suppress the noise while reducing speech distortion. Experimental results, including several objective measurements, show that the proposed wavelet-based algorithm outperforms spectral subtraction and other wavelet-based denoising approaches for speech enhancement for nonstationary noise environments.

In this paper, we propose a new mask estimation method for the computational auditory scene analysis (CASA) of speech using two microphones. The proposed method is based on a hidden Markov model (HMM) in order to incorporate an observation that the mask information should be correlated over contiguous analysis frames. In other words, HMM is used to estimate the mask information represented as the interaural time difference (ITD) and the interaural level difference (ILD) of two channel signals, and the estimated mask information is finally employed in the separation of desired speech from noisy speech. To show the effectiveness of the proposed mask estimation, we then compare the performance of the proposed method with that of a Gaussian kernel-based estimation method in terms of the performance of speech recognition. As a result, the proposed HMM-based mask estimation method provided an average word error rate reduction of 61.4% when compared with the Gaussian kernel-based mask estimation method.

A transceiver macro for high-speed data transmission via cable in vehicles is proposed. The transceiver uses ac coupling and bi-directional interface topology for protecting LSIs against unexpected short of cable and harness/chassis and has a spread-spectrum-clocking (SSC) generator that reduces noise due to electromagnetic interference. A driver current control has been used for fast switching of data direction on ac-coupled interfaces. An adaptive bandwidth control has been used in a Δ ∑ PLL to improve SCC significantly. A test chip has been fabricated and shows stable and bi-directional data communication with data rate of 162 to 972 Mbps through 20-m cable. Thanks to an optimum calibration of the SSC-PLL bandwidth, it reduces peak power at 33 kHz by -23 dB and provides 2% modulation at a data rate of 810 Mbps.

This paper proposes a low-voltage high-speed sample-and-hold (S/H) structure with excellent power efficiency. Based on the switched-opamp technique, an inverse-flip-around architecture which maximizes the feedback factor is employed in the proposed S/H. A skew-insensitive double-sampling mechanism is presented to increase the throughput by a factor of two while eliminating the timing mismatch associated with double-sampling circuits. Furthermore, a dual-input dual-output opamp is proposed to incorporate double-sampling into the switched-opamp based S/H. This opamp also removes the memory effect in double-sampling circuitry and features fast turn-on time to improve the speed performance in switched-opamp circuits. Simulation results using a 0.13-µm CMOS process model demonstrates the proposed S/H circuit has a total-harmonic-distortion of -67.3 dB up to 250 MSample/s and a 0.8 VPP input range at 0.8 V supply. The power consumption is 3.5 mW and the figure-of-merit is only 7.4 fJ/step.

In this paper, we propose a method for designing a set of pulses whose spectrum is efficiently contained in amplitude and bandwidth. Because these pulses are derived from and have shapes that are either equal or similar to the Hermite pulses, we name our proposed transmit pulses as spectrally efficient Hermite pulses. Given that the proposed set of pulses does not constitute an orthonormal one, we also propose a set of receive templates which permit orthonormal detection of the incoming signals at the receiver. The importance of our proposal is in the potential implementation of M-ary pulse shape modulation systems, for ultra wideband communications, with sets of pulses that are efficiently contained within a specific bandwidth and limited to a certain amplitude.

The spectral efficiency of cooperative relaying in interference-limited environments in which a given channel is spatially reused is investigated. Cooperative relaying is a promising technique that uses neighboring stations to forward the data toward the destination in order to achieve spatial diversity gain. It has been reported that by introducing cooperative relaying into communication between an isolated source-destination pair, the error rate or spectral efficiency is generally improved. However, it is not intuitively clear whether cooperative relaying can improve the performance in interference-limited environments because the simultaneous transmission of multiple stations increases the number of interference signals. Assuming the most fundamental cooperative relaying arrangement, which consists of only one relay station, numerical results reveal that cooperative relaying is not always superior to non-cooperative single-hop and two-hop transmissions in terms of spectral efficiency.

In this letter, we propose a new histogram equalization method to compensate for acoustic mismatches mainly caused by corruption of additive noise and channel distortion in speech recognition. The proposed method employs an improved test cumulative distribution function (CDF) by more accurately smoothing the conventional order statistics-based test CDF with the use of window functions for robust feature compensation. Experiments on the AURORA 2 framework confirmed that the proposed method is effective in compensating speech recognition features by reducing the averaged relative error by 13.12% over the order statistics-based conventional histogram equalization method and by 58.02% over the mel-cepstral-based features for the three test sets.

Break arcs are generated in a DC 42 V-10 A resistive circuit. The contact material is Ag or Ag/ZnO. The number of break operations is two hundreds for each contact material. The motion of break arcs is observed with a high-speed camera. Relationship between the dependence of arc duration on the number of operations and the motion of arc spots is investigated. The following results are shown. For Ag contacts the arc duration is almost constant independent to the number of break operations. For Ag/ZnO contacts, on the other hand, the arc duration changes irregularly to short (59 ms) or long (69 ms) arc-duration after 30th break operation. The moving range of arc spots on contact surfaces is broad for the case of short arc-duration and is narrow for the case of long arc-duration. The cause of the results for Ag/ZnO contacts is considered that the difference of the boiling points of Ag and ZnO leads to the porous structure on the contact surface.

In a D.C.42 V-10A resistive circuit, break arcs are generated between electrical contact pairs. The materials of the contact pairs are Ag, Ag/C 2wt%, Ag/SnO2 12wt%, and Ag/ZnO 12wt%. The arc spectral intensities are measured by a time-resolved spectroscopic temperature measurement system. The arc temperature is calculated from the spectral intensities by using the method of relative intensities of two spectra. The experimental results are as follows. The arc temperature gradually decreases with increase of the gap of electrical contacts. The ranges of arc temperature for Ag, Ag/C 2wt%, Ag/SnO2 12wt%, and Ag/ZnO 12wt% contacts pairs are 4500-11000 K, 4000-6000 K, 4000-7000 K, and 4000-11000 K, respectively.

At high signal frequencies (i.e. in the GHz range), a connector must be considered as part of an electromagnetic transmission line. At these frequencies, the impedance characteristics of the connector stemming from the distributed inductance and capacitance of pins and the associated wiring, must be carefully controlled; insertion losses must be minimized and undesirable coupling between non-neighboring pins giving rise to crosstalk must be avoided to achieve optimal signal transmission. This paper reviews fundamental issues associated with the performance optimization of multi-conductor connector structures for high speed signal transmission. The paper complements an earlier publication that reviewed the major factors affecting electrical contact resistance at high frequencies [1].

Motion of break arcs driven by external magnetic field is observed using a high-speed camera. The magnetic field is applied with a permanent magnet. Experimental circuit is DC42 V-10 A resistive circuit. Material of electrical contacts is silver. Following results are shown. The break arcs are driven in the direction according to Lorentz force. The arc duration decreases with decrease of the distance between the electrical contacts and the magnet. When the external magnetic-flux density at the position of the break arc is lower than a certain value, the effect of the magnetic field to drive the break arc becomes ineffective to shorten the arc duration. The result is explained with a relationship between the motion of break arc and the distribution of the external magnetic field.

Most of previous works have presented the dynamic spectrum allocation (DSA) gain achieved by utilizing the time or regional variations in traffic demand between multi-network operators (NOs). In this paper, we introduce the functionalities required for the entities related with the spectrum sharing and allocation and propose a spectrum allocation algorithm while considering the long-term priority between NOs, the priority between multiple class services, and the urgent bandwidth request. To take into account the priorities among the NOs and the priorities of multiple class services, a spectrum sharing metric (SSM) is proposed, while a negotiation procedure is proposed to treat the urgent bandwidth request.

Interleaved Frequency Division Multiplexing (IFDM) can achieve high diversity gain as well as establishing orthogonal frequency multiplexing by using a comb-shaped frequency spectrum. In IFDM, as the number of repeat transmissions of a modulated symbol is increased, the comb-shaped frequency spectrum should be narrowed, so that the frequency diversity gain is decreased. In addition, IFDM suffers from inter-path interference imposed on the transmitted signal by multipath fading channel. In this paper, a novel frequency spectrum construction is proposed. In the proposed frequency spectrum construction, the comb-shaped frequency spectrum is frequency-shifted for every modulated symbol. As a result, the frequency spectrum of the frame composed of many modulated symbols is widely spread. In addition, the inter-path interference can be suppressed because the modulated symbol is orthogonal, in the frequency domain, to the following symbol. From the computer simulation, the frequency spectrum rotation can achieve better error rate performance thanks to the increase in frequency diversity gain and suppressing inter-path interference.

The adaptive predistorter and the negative feedback system are known as methods to compensate for the nonlinear distortion of a power amplifier. Although the feedback method is a simple technique, its instability impedes the capability of high-feedback gain to achieve a high-compensation effect. On the other hand, the predistorter requires a long time for convergence of the adaptive predistorters. In this paper, we propose a nonlinear distortion compensation method for a narrow-band signal. In this method, an adaptive predistorter and negative feedback are combined. In addition, to shorten the convergence time to minimize nonlinear distortion, a variable step-size (VS) method is also applied to the algorithm to determine the parameters of the adaptive predistorter. Using computer simulations, we show that the proposed scheme achieves both five times faster convergence speed than that of the predistorter and three times higher permissible delay time in the feedback amplifier than that of a negative feedback only amplifier.