We propose a new parameter-extraction approach based on a mixed-mode genetic algorithm (GA), including the efficient search-space separation and local-minima-convergence prevention process. The technique, substantially extended from our previous work, allows the designed figures-of-merit, such as internal quantum efficiency (ηi) as well as transparency current density (Jtr) of lasers and minimum noise figure (NFmin) as well as associated available gain (GA,assoc) of low-noise amplifiers (LNAs), extracted by an analytical equation-based methodology combined with an evolutionary numerical tool. Extraction results, which agree well with actually measured data, for both state-of-the-art InGaAs quantum-well lasers and advanced SiGe LNAs are presented for the first time to demonstrate this multi-parameter analysis and high-accuracy optimization.

Estimating the ratio of two probability density functions (a.k.a. the importance) has recently gathered a great deal of attention since importance estimators can be used for solving various machine learning and data mining problems. In this paper, we propose a new importance estimation method using a mixture of probabilistic principal component analyzers. The proposed method is more flexible than existing approaches, and is expected to work well when the target importance function is correlated and rank-deficient. Through experiments, we illustrate the validity of the proposed approach.

We have developed a one-to-many eigenvoice conversion (EVC) system that allows us to convert a single source speaker's voice into an arbitrary target speaker's voice using an eigenvoice Gaussian mixture model (EV-GMM). This system is capable of effectively building a conversion model for an arbitrary target speaker by adapting the EV-GMM using only a small amount of speech data uttered by the target speaker in a text-independent manner. However, the conversion performance is still insufficient for the following reasons: 1) the excitation signal is not precisely modeled; 2) the oversmoothing of the converted spectrum causes muffled sounds in converted speech; and 3) the conversion model is affected by redundant acoustic variations among a lot of pre-stored target speakers used for building the EV-GMM. In order to address these problems, we apply the following promising techniques to one-to-many EVC: 1) mixed excitation; 2) a conversion algorithm considering global variance; and 3) adaptive training of the EV-GMM. The experimental results demonstrate that the conversion performance of one-to-many EVC is significantly improved by integrating all of these techniques into the one-to-many EVC system.

In this paper, we propose an acoustic model that is robust to multiple noise environments, as well as a method for adapting the acoustic model to an environment to improve the model. The model is called "the multi-mixture model," which is based on a mixture of different HMMs each of which is trained using speech under different noise conditions. Speech recognition experiments showed that the proposed model performs better than the conventional multi-condition model. The method for adaptation is based on the aspect model, which is a "mixture-of-mixture" model. To realize adaptation using extremely small amount of adaptation data (i.e., a few seconds), we train a small number of mixture models, which can be interpreted as models for "clusters" of noise environments. Then, the models are mixed using weights, which are determined according to the adaptation data. The experimental results showed that the adaptation based on the aspect model improved the word accuracy in a heavy noise environment and showed no performance deterioration for all noise conditions, while the conventional methods either did not improve the performance or showed both improvement and degradation of recognition performance according to noise conditions.

In this paper, we describe a novel model training method for one-to-many eigenvoice conversion (EVC). One-to-many EVC is a technique for converting a specific source speaker's voice into an arbitrary target speaker's voice. An eigenvoice Gaussian mixture model (EV-GMM) is trained in advance using multiple parallel data sets consisting of utterance-pairs of the source speaker and many pre-stored target speakers. The EV-GMM can be adapted to new target speakers using only a few of their arbitrary utterances by estimating a small number of adaptive parameters. In the adaptation process, several parameters of the EV-GMM to be fixed for different target speakers strongly affect the conversion performance of the adapted model. In order to improve the conversion performance in one-to-many EVC, we propose an adaptive training method of the EV-GMM. In the proposed training method, both the fixed parameters and the adaptive parameters are optimized by maximizing a total likelihood function of the EV-GMMs adapted to individual pre-stored target speakers. We conducted objective and subjective evaluations to demonstrate the effectiveness of the proposed training method. The experimental results show that the proposed adaptive training yields significant quality improvements in the converted speech.

Chip-to-chip serial data communication is superposed on power supply over common Vdd/Vss connections through chip, package, and board traces. A power line transceiver demonstrates half duplex spiking communication at more than 100 Mbps. A pair of transceivers consumes 1.35 mA from 3.3 V, at 130 Mbps. On-chip power line LC low pass filter attenuates pseudo-differential communication spikes by 30 dB, purifying power supply current for internal circuits. Bi-directional spiking communication was successfully examined in a 90-nm CMOS prototype setup of on-chip waveform capturing. A micro controller forwards clock pulses to and receives data streams from a comparator based waveform capturer formed on a different chip, through a single pair of power and ground traces. The bit error rate is small enough not to degrade waveform acquisition capability, maintaining the spurious free dynamic range of higher than 50 dB.

Bag-of-Visual-Words representation has recently become popular for scene classification. However, learning the visual words in an unsupervised manner suffers from the problem when faced these patches with similar appearances corresponding to distinct semantic concepts. This paper proposes a novel supervised learning framework, which aims at taking full advantage of label information to address the problem. Specifically, the Gaussian Mixture Modeling (GMM) is firstly applied to obtain "semantic interpretation" of patches using scene labels. Each scene induces a probability density on the low-level visual features space, and patches are represented as vectors of posterior scene semantic concepts probabilities. And then the Information Bottleneck (IB) algorithm is introduce to cluster the patches into "visual words" via a supervised manner, from the perspective of semantic interpretations. Such operation can maximize the semantic information of the visual words. Once obtained the visual words, the appearing frequency of the corresponding visual words in a given image forms a histogram, which can be subsequently used in the scene categorization task via the Support Vector Machine (SVM) classifier. Experiments on a challenging dataset show that the proposed visual words better perform scene classification task than most existing methods.

A new mixed-voltage I/O buffer for low-voltage low-latency operation is proposed in this paper. The proposed buffer adopts a novel delay-based timing-control scheme to efficiently avoid problems like gate-oxide stress and hot-carrier degradation. The proposed timing-control scheme also allows the buffer to have a lower latency for transmitting data by avoiding the use of timing-critical circuits like series-connected transmission gates (TGs) and triple-stacked transistors. The latency for receiving data at low supply voltage is also reduced by employing a variable stacked transistor gate-biasing scheme. Comparison results in an 80-nm CMOS process indicated that the proposed mixed-voltage I/O buffer improved up to 79.3% for receiving the external data and up to 23.8% for transmitting the internal data at a supply voltage of 1.2 V.

A low-voltage quadrature up-conversion CMOS mixer for 5-GHz wireless communication applications is designed with a TSMC 0.18-µm process. The fold-switching technique is used to implement the low-voltage double balanced quadrature mixer. A miniature lumped-element microwave broadband rat-race hybrid and RLC shift network are used for the local oscillator and the intermediate frequency port design, respectively. The measured results demonstrate that the mixer can reach a high conversion gain, a low noise figure (NF), and a high linearity. The mixer exhibits improvement in noise, conversion gain, and image rejection. The mixer shows a conversion gain of 16 dB, a noise figure of 12.8 dB, an image rejection of 45 dB, while dissipating 15.5 mW for an operating voltage at 1 V.

Low noise terahertz (THz) receivers based on superconducting niobium nitride (NbN) hot electron bolometer (HEB) mixers have been designed, fabricated and measured for applications in astronomy and cosmology. The NbN HEB mixer consists of a planar antenna and an NbN bridge connecting across the antenna's inner terminals on a high-resistivity Si substrate. To eliminate the influence of direct detection and instability of the local oscillation (LO) power, a wire grid has been used to change the input LO power for compensating the shift of bias current during Y-factor measurement. The double sideband (DSB) receiver noise temperatures at 4.2 K without corrections have been measured from 0.65 to 3.1 THz. The excess quantum noise factor β of about 4 has been obtained, which agrees well with the calculated value. Allan variance of the HEB has been characterized, and Allan time TA longer than 0.4 s is obtained. We also estimated the temperature resolution of the HEB from the Allan variance and obtained the minimum temperature resolution of 1.1 K using a Gunn oscillator with its multipliers at 0.65 THz as an LO source.

Carbon nanotubes (CNTs) have generated great interest within the many areas of nanotechnology due to their superior and outstanding physical properties. However effective dispersion in many solvents has imposed limitations upon the use of CNTs in a number of novel applications. Functionalization presents a solution for CNTs to be more soluble which make them integrate well into any organic, inorganic or biological systems. CNTs can be easily functionalized using cyclodextrin (CD) treatment. The CD modification of carbon nanotubes is both simple and effective. It requires no prolonged heating, filtration and washing which can severely damage the small diameter nanotubes. The formation of surface functional groups and changes of nanotubes structures of functionalized carbon nanotubes (f-CNTs) were monitored by Fourier transform infrared spectroscopy (FTIR), Thermo gravimetric analysis (TGA) and field emission scanning electron microscopy (FESEM), respectively. From the TGA results, the amount of weight loss of the f-CNTs in varying ratios indicated the amount of CD that was functionalized. It was also noted that the FTIR spectra showed the presence of functional groups associated with CD in the f-CNTs. As a result, the cyclodextrin groups were found to be possibly adsorbed at the surface of the nanotubes walls. The f-CNTs showed substantial solubility in N-methyl-2-pyrrolidone (NMP) which helps in a better distribution of the CNTs in the mixed matrix membrane (MMM) prepared. Hence, the influence of the f-CNTs in the polymer matrix will give rise to enhanced physical properties of the MMM suitable for applications in gas separations.

A 0.4-5.8 GHz SiGe-MMIC quadrature modulator (Q-MOD) employing a self current controlled mixer for cognitive radio is described. The self current controlled mixer consists of a Gilbert cell mixer and a self current control circuit which is composed of both a current feedback circuit and an output buffer amplifier. The self current control circuit automatically controls the mixer current according to the output power level, and improves the linearity over wide radio frequency (RF) range. Simulation results show that the proposed Q-MOD realizes 1 dB compression point (P1 dB) improvement of more than 3.0 dB compared to the conventional Q-MOD at the frequencies of 0.4, 0.8, 1.95, 5.2 and 5.8 GHz. The fabricated Q-MOD achieves P1 dB improvement of more than 2.8 dB under the same condition. It also improves the output power with error vector magnitude (EVM) of 3.0% (Pout@EVM=3.0%), and achieves the Pout improvement of more than 2.7 dB under the modulation conditions of UHF wireless system (OFDM/16QAM, 0.4 GHz), W-CDMA (HPSK/QPSK, 0.8 GHz/1.95 GHz) and wireless-LAN (OFDM/64QAM, 5.2 GHz/5.8 GHz).

This paper presents a technique for class-dependent decoding for statistical machine translation (SMT). The approach differs from previous methods of class-dependent translation in that the class-dependent forms of all models are integrated directly into the decoding process. We employ probabilistic mixture weights between models that can change dynamically on a sentence-by-sentence basis depending on the characteristics of the source sentence. The effectiveness of this approach is demonstrated by evaluating its performance on travel conversation data. We used this approach to tackle the translation of questions and declarative sentences using class-dependent models. To achieve this, our system integrated two sets of models specifically built to deal with sentences that fall into one of two classes of dialog sentence: questions and declarations, with a third set of models built with all of the data to handle the general case. The technique was thoroughly evaluated on data from 16 language pairs using 6 machine translation evaluation metrics. We found the results were corpus-dependent, but in most cases our system was able to improve translation performance, and for some languages the improvements were substantial.

A 2 to 5 GHz-band self frequency dividing quadrature mixer utilizing current re-use configuration with small size and broad band operation is proposed for a direct conversion receiver and a low-IF receiver of cognitive radio. The proposed mixer operates at twice the LO frequency by directly using a static type flip-flop frequency divider as the LO switching circuit for quadrature signal generation. The current re-use configuration is realized because the dc current of the frequency divider and the RF common-emitter amplifier share the same current flow path. Simulations and experiments verify that the proposed mixer offers broad band operation, miniaturization, and low power consumption. The mixer IC fabricated by 0.35 µm SiGe-BiCMOS technology achieved the conversion gain of 20.6 dB, noise figure of 11.9 dB and EVM for π/4-shift QPSK signal of 4.4% at 2.1 GHz with power consumption of 15 mW and size of 0.22 0.31 mm2. For the confirmation of broad band operation, the characteristics of conversion gain and noise figure were measured at 5.2 GHz. The proposed mixer could operate at 5.2 GHz with enough conversion gain, but the noise figure was inferior to that of 2.1 GHz. Therefore the further investigation and improvement about the noise figure will be needed for higher frequency.

The ratio of two probability densities is called the importance and its estimation has gathered a great deal of attention these days since the importance can be used for various data processing purposes. In this paper, we propose a new importance estimation method using Gaussian mixture models (GMMs). Our method is an extention of the Kullback-Leibler importance estimation procedure (KLIEP), an importance estimation method using linear or kernel models. An advantage of GMMs is that covariance matrices can also be learned through an expectation-maximization procedure, so the proposed method--which we call the Gaussian mixture KLIEP (GM-KLIEP)--is expected to work well when the true importance function has high correlation. Through experiments, we show the validity of the proposed approach.

This paper presents the mixed time-frequency steady-state analysis method for efficient simulation of circuits whose excitation frequencies are widely separated. These circuits can be written by multitime partial differential equations. In this paper, an axis of the slow time-scale is formulated in the time domain and another axis of the fast time-scale is formulated in the frequency domain. We show that computational cost, however, is not dependent on the interval of frequencies, whereas for the harmonic balance or transient analysis, it increases as the interval of frequencies increases.

A low-power K-band CMOS current-mode up-conversion mixer is proposed. The proposed mixer is realized using four analog current-squaring circuits. This current-mode up-conversion mixer is fabricated in 0.13-µm 1P8M triple-well CMOS process, and has the measured power conversion gain of -5 dB. The fabricated CMOS up-conversion mixer dissipates only 3.1 mW from a 1-V supply voltage. The VCO can be tuned from 20.8 GHz to 22.7 GHz. Its phase noise is -108 dBc/Hz at 10-MHz offset frequency. It is shown that the proposed mixer has great potential for low-voltage and low-power CMOS transmitter front-ends in advanced nano-CMOS technologies.

In this paper, an image restoration method using the Wiener filter is proposed. In order to bring the theory of the Wiener filter consistent with images that have spatially varying statistics, the proposed method adopts the locally adaptive Wiener filter (AWF) based on the universal Gaussian mixture distribution model (UNI-GMM) previously proposed for denoising. Applying the UNI-GMM-AWF for deconvolution problem, the proposed method employs the stationary Wiener filter (SWF) as a pre-filter. The SWF in the discrete cosine transform domain shrinks the blur point spread function and facilitates the modeling and filtering at the proceeding AWF. The SWF and UNI-GMM are learned using a generic training image set and the proposed method is tuned toward the image set. Simulation results are presented to demonstrate the effectiveness of the proposed method.

A 0.13-µm CMOS 2.4-GHz low-noise balun-mixer is proposed, where a noise-canceling transconductance stage is adopted for low-noise characteristics. A current-bleeding circuit with an LC resonator is also adopted to further improve the noise figure of the proposed balun-mixer, without additional DC power consumption. The measured results show a DSB NF of 5.5 dB over output IF frequency ranges of 10 to 100 MHz, a conversion gain of 19 dB, and an input P1 dB of -16 dBm. The proposed balun-mixer is implemented in 0.13-µm CMOS technology and consumes only 4.5 mA from a 1.5-V supply voltage.

In this paper, the research advances in SIW-like (Substrate Integrated Waveguide-like) guided wave structures and their applications in the State Key Laboratory of Millimeter Waves of China is reviewed. Our work is concerned with the investigations on the propagation characteristics of SIW, half-mode SIW (HMSIW) and the folded HMSIW (FHMSIW) as well as their applications in microwave and millimeter wave filters, diplexers, directional couplers, power dividers, antennas, power combiners, phase shifters and mixers etc. Selected results are presented to show the interesting features and advantages of those new techniques.