Speech recognition in living rooms: Integrated speech enhancement and recognition system based on spatial,spectral and temporal modeling of sounds

Research on noise robust speech recognition has mainly focused on dealing with relatively stationary noise that may differ from the noise conditions in most living environments. In this paper, we introduce a recognition system that can recognize speech in the presence of multiple rapidly time-varying noise sources as found in a typical family living room. To deal with such severe noise conditions, our recognition system exploits all available information about speech and noise; that is spatial (directional), spectral and temporal information. This is realized with a model-based speech enhancement pre-processor, which consists of two complementary elements, a multi-channel speech–noise separation method that exploits spatial and spectral information, followed by a single channel enhancement algorithm that uses the long-term temporal characteristics of speech obtained from clean speech examples. Moreover, to compensate for any mismatch that may remain between the enhanced speech and the acoustic model, our system employs an adaptation technique that combines conventional maximum likelihood linear regression with the dynamic adaptive compensation of the variance of the Gaussians of the acoustic model. Our proposed system approaches human performance levels by greatly improving the audible quality of speech and substantially improving the keyword recognition accuracy.     

Sensorless vector controlled converter for variable speed wind generation system using an induction generator

This paper proposed two types of speed sensorless converters that can control rotational speed and electric power. One is based on a slip frequency control system. The other is based on a vector control system. The rotational speed of the wind turbine is estimated with the phase voltage and phase current of the induction generator by the adaptive rotor flux observer. The estimated wind turbine rotational speed ω_r^est is used as the feedback of the speed control loop in the converter control system. Also, the estimated rotor flux ?_2d^est is used for the vector control. The simulation results confirm that both of them perform satisfactorily under the speed sensorless operation. The method based on the vector control system generates more electrical energy than does the method based on the slip frequency control system. © 2007 Wiley Periodicals, Inc. Electr Eng Jpn, 159(4): 62–75, 2007; Published online in Wiley InterScience ( www.interscience.wiley.com ). DOI 10.1002/eej.20323     

Structure Analysis of Molten Ba–Ge Alloys Using Electrostatic Levitation Technique Combined With High-Energy X-Ray Diffraction

High-energy synchrotron X-ray diffraction experiments were performed for clathrate-forming barium–germanium (Ba–Ge) alloys in the liquid state near the eutectic composition. The accurate structure information of highly reactive melts has been obtained by applying electrostatic levitation technique as a containerless method. The structure information obtained from the reverse Monte Carlo simulation suggests that the short-range ordering between the Ge atoms occurs with the addition of Ba atoms in the liquid Ba–Ge alloys. This can be associated with the formation of the cage-like structure composed of the Ge atoms in the liquid alloys near the eutectic composition.     

Potentiality of a new minimum‐order back‐EMF state observer taking acceleration into account for sensorless vector control of permanent‐magnet synchronous motors

This paper proposes new sensorless vector control methods for permanent‐magnet synchronous motors (PMSMs), which are based on a new minimum‐order back‐EMF state observer taking acceleration into account, and examines potentiality of the improved back‐EMF observer through experiments. Conventional sensorless vector control methods for PMSMs using an estimate of back‐EMF are established under the assumption of constant speed or zero acceleration, and consequently cannot operate properly at modes requiring rapid speed change especially in low speed region. On the other hand, the proposed back‐EMF observer has the following features: (1) it is a new back‐EMF state observer taking acceleration into account and requiring no additional approximation to motor mathematical model; (2) it is a minimum‐order state observer; (3) it utilizes motor parameters in the simplest manner; (4) it can be applied to both salient‐pole and non‐salient‐pole PMSMs; (5) it can be realized in both rotor and stator reference frames. Detailed designs and analyses for the improved state observer and the sensorless vector control systems in both rotor and stator reference frames are given. In relation to the sensorless vector control systems, this paper newly proposes a double‐integral‐type PLL method and an integral‐feedback‐type acceleration/speed estimation method. Their potentialities are examined through experiments. © 2008 Wiley Periodicals, Inc. Electr Eng Jpn, 164(4): 78–92, 2008; Published online in Wiley InterScience ( www.interscience.wiley.com ). DOI 10.1002/ eej.20630     

Photoelectrochemical decomposition of bio-related compounds at a nanoporous semiconductor film photoanode and their photocurrent-photovoltage characteristics

Photoelectrochemical decomposition of bio-related compounds such as amino acids was investigated with a biophotochemical cell comprising a mesoporous TiO₂ thin film photoanode and an O₂-reducing cathode. It was concluded that a kind of Schottky junction formed at the surface of the TiO₂ (called as liquid junction) induced the photodecomposition followed by generation of photocurrent/photovoltage. Complete photodecomposition was investigated by the CO₂ formation yield. The photocurrent-photovoltage (J-V) characteristics of amino acids and other typical bio-related compounds were investigated, and the short circuit photocurrent (J_sc), open circuit photovoltage (V_oc), and Fill factor (ff) were exhibited. Effect of pH on the photodecomposition of phenylalanine and cysteine were studied; for cysteine alkaline conditions gave a high efficiency, which was interpreted by the high electron-donating ability of the dissociated -S⁻ group. The incident light-to-current conversion efficiency (IPCE) of cysteine was 25% at 350 nm. It was for the first time shown that organic acids gave high internal quantum efficiency (η′) over 8 (=800%) in the photodecomposition; for oxalic acid it was 9.3 (=930%) and for butyric acid 8.2. The alternating current impedance spectroscopy of glycine showed that the cell performance is determined by the chemical reactions at TiO₂ or Pt electrodes.     

A New Nonmetallic Inclusion Rating Method by Positive Use of Hydrogen Embrittlement Phenomenon

A new inclusions rating method using hydrogen embrittlement of a tensile test specimen is proposed. This method is essentially based on the statistics of extremes for inclusion rating where the maximum inclusion size is determined by simple tensile testing of a hydrogen-precharged (H-precharged) specimen. Tensile tests were conducted using two bearing steels (SAE52100 HV 346, HV 447, HV 559, HV 611, HV 678 and ASTM-A485-1 HV 706, HV 715) and one spring steel (SAE5160, HV 651). Fatigue tests were conducted using SAE52100 bearing steel (HV 682). All H-precharged tensile specimens (SAE52100, ASTM-A485-1 and SAE5160) were fractured from internal inclusions except the SAE52100 tensile specimens with a Vickers hardness of HV 346. It was confirmed that the distribution of extreme values of inclusion sizes obtained by SAE52100 tensile testing with H-precharged specimens coincided with those obtained by SAE52100 fatigue testing. From these results, it is presumed that the inclusion rating method by fatigue testing can be replaced by simple tensile testing with H-precharged specimens. The proposed method is more convenient and reliable than other existing inclusion rating methods, i.e., fatigue testing and optical microscopy. The proposed method can be applied to specimens with a Vickers hardness of higher than HV 447.     

Annealing effect on temperature coefficient of resistivity in La1−xSrxMnO3 ceramics

We investigated annealing effects of La_1?xSr_xMnO₃ (x = 0–0.6) on electrical resistivity and the temperature coefficient of resistivity (TCR). The annealed samples’ resistivity was lower than those of non-annealed samples. For example, annealing changed the resistivity of x = 0.3 at 25 °C from 4.50 × 10^?5 to 3.71 × 10^?5 Ω m. Remarkable difference in TCR was observed after annealing, for x = 0.3, 0.45, and 0.5. For x = 0.3, the TCR after annealing was 4000 ppm/°C, which was 1250 ppm/°C greater than that before annealing. We investigated (1) crystal phase, (2) Mn average valence, (3) Mott insulator–metal transition temperature, and (4) microstructure. The microstructure was remarkably varied for annealed x = 0.3 and 0.5. The average grain size of the x = 0.3 increased from 1.60 up to 2.38 μm. Results show that annealing affects resistivity and TCR because of grain growth during annealing.     

Hybridized synergy effect among TiO2, Pt and graphite silica on photocatalytic hydrogen production from water–methanol solution

The photocatalytic production of hydrogen gas from aqueous methanol solution was performed using powdered mixtures of graphite silica (GS) and platinized TiO₂ (Pt/TiO₂) or those of GS, Pt and TiO₂. The addition of GS to Pt/TiO₂ resulted in the decrease of the amount of H₂ gas, whereas the addition of GS to mixtures of TiO₂ and Pt led to the incremental production of H₂ gas. This effect is attributed to the aggregation of GS, TiO₂ and Pt in suspension. The new additional electron transfer process of TiO₂ → Pt → GS caused the increment of the amount of hydrogen gas.     

To develop a quantitative method for predicting shrinkage porosity in squeeze casting

In order to secure high strength and high elongation of suspension parts, it is critical to predict shrinkage porosity quantitatively. A new simulation method for quantitative predic'don of shrinkage porosity when replenishing molten metal has been proposed for squeeze casting process. To examine the accuracy of the calculation model, the proposed method was applied to a plate model.