Multi-stream LSTM-HMM decoding and histogram equalization for noise robust keyword spotting

Highly spontaneous, conversational, and potentially emotional and noisy speech is known to be a challenge for today’s automatic speech recognition (ASR) systems, which highlights the need for advanced algorithms that improve speech features and models. Histogram Equalization is an efficient method to reduce the mismatch between clean and noisy conditions by normalizing all moments of the probability distribution of the feature vector components. In this article, we propose to combine histogram equalization and multi-condition training for robust keyword detection in noisy speech. To better cope with conversational speaking styles, we show how contextual information can be effectively exploited in a multi-stream ASR framework that dynamically models context-sensitive phoneme estimates generated by a long short-term memory neural network. The proposed techniques are evaluated on the SEMAINE database—a corpus containing emotionally colored conversations with a cognitive system for “Sensitive Artificial Listening”.     

黔中喀斯特山区村域稻田土壤理化特性的空间变异特征及空间自相关性

以贵州省安顺市西秀区旧州镇文星村为研究区域,采用地统计学方法,研究了黔中喀斯特山区土壤容重、总孔隙度、毛管持水量、毛管孔隙度、通气孔隙度等物理特性与土壤pH、有机质、速效养分(S、Si、N、P、K、Fe、Zn、Cu、Mn)、全量养分(N、P、K)、交换性盐(Ca、Mg)及阳离子交换量等化学特性的空间变异特征及自相关性。在土壤各特征值中,pH、毛管孔隙度变异系数分别为4.59%、8.28%、9.83%,为弱变异;其他变异系数在10%-100%之间,为中等程度变异。半方差分析表明,土壤毛管持水量、通气孔隙度、毛管孔隙度、速效养分(S、Si、Cu、Mn)、全量养分(N、P、K)、交换性盐(Ca、Mg)及阳离子交换量的C₀/(C₀+C)<25%,表现为强烈的空间自相关性;容重、pH、速效养分(N、P、K、Fe、Zn)及有机质的C₀/(C₀+C)在25%-75%之间,表现为中等空间自相关性,而Moran''s I指数分析表明,除毛管孔隙度、有效S、速效P及有效Mn的空间自相关性较弱外,其他均呈正的显著空间自相关性。根据克里金插值图,在研究区域内土壤毛管孔隙度及毛管持水量与pH、速效养分(S、Si、Zn、Cu、)全P、全N、有机质、交换性Ca、Mg及阳离子交换量在东北方向和西北角分布较低,在西南角方向土壤容重及通气孔隙度分布较低,而pH、速效K、有效Cu、全P、全K、有机质、阳离子交换量分布较高,且速效养分(N、K、S、Cu)、全P、有机质的含量及阳离子交换量均存在从西南向东北方向递减的趋势。     

Improvement of the muscle fractional multimodel for low-rate stimulation

Modeling of isometric contractions, due to motor unit (MU) stimulations of Peroneus digiti quarti and peroneus brevis rat muscles, is presented. The modeling is realized through a multimodel, which allows distinguishing asymmetric contractions and relaxation mechanisms for time domain identification. First, this paper compares two fractional functions and a rational transfer function that model each phase of IIA and IIB MU twitch. The advantages of using fractional functions are underlined, since the number of parameters are minimized. Indeed, fractional models, due to its infinite dimension nature, are particularly adapted not only to model complex systems with few parameters but also to obtain a real time exploitable model for a salamander robot simulator. Finally, muscle response for 10 Hz pulse stimulation shows non-stationary characteristics. A method, modeling the transient state of muscle responses and introducing time varying parameters, is presented.     

Statistical properties of the dichotomous noise generated in biochemical processes

Dichotomous noise detected with the help of various single-molecule techniques convincingly reveals the actual occurrence of a multitude of conformational substates composing the native state of proteins. The nature of the stochastic dynamics of transitions between these substates is determined by the particular statistical properties of the noise observed. These involve nonexponential and possibly oscillatory time decay of the second order autocorrelation function, its relation to the third order autocorrelation function, and a relationship to dwell-time distribution densities and their correlations. Processes gated by specific conformational substates are distinguished from those with fluctuating barriers. This study throws light on the intriguing matter of the possibility of multiple stepping of the myosin motor along the actin filament per ATP molecule hydrolyzed. Paper authored by participants of the international conference: International Workshop on Ionic Channels, Szczyrk, Poland, May 27 – June 01, 2007. Publication cost was covered by the organisers of this meeting.     

Improved fractional precipitation method for purification of paclitaxel

This study investigated changing the methanol/water ratio during fractional precipitation of paclitaxel, and adding all the distilled water at room temperature, followed mixing for an additional 10 min. When the methanol/water ratio was 50:50, 40:60, and 30:70 (v/v), the paclitaxel yield was 42.0%, 84.3%, and 92.0%, respectively. When using a methanol/water ratio of 50:50 (v/v), a similar high purity and yield of paclitaxel to the case of storing at a low temperature was achieved when adding all the distilled water at room temperature, followed by additional mixing for 10 min and further mixing at room temperature during fractional precipitation. Thus, additional mixing after adding all the distilled water is confirmed as important during fractional precipitation. Furthermore, the present results show that a high yield of high-purity paclitaxel is possible with additional mixing at room temperature after adding all the distilled water, which is significantly more economical than the existing method of storing at a low temperature for a long time after adding all the distilled water during fractional precipitation.     

Recent advances in DNA sequencing methods - general principles of sample preparation

DNA sequencing has revolutionized biomedicine, and progress in the field has been unrelenting since it was invented over 30 years ago. The complete DNA sequence of the human genome was obtained as the culmination of a decade of work by a large number of scientists. Less than ten years later, so-called ‘next-generation’ instruments now make it possible for a single lab to produce the same amount of data in a week. But while the instruments are increasingly automated, upstream sample processing remains a challenge. Here I review the current state of the art in preparing genomic and RNA samples for high throughput sequencing.     

Modification of sample size in group sequential clinical trials

In group sequential clinical trials, sample size reestimation can be a complicated issue when it allows for change of sample size to be influenced by an observed sample path. Our simulation studies show that increasing sample size based on an interim estimate of the treatment difference can substantially inflate the probability of type I error in most practical situations. A new group sequential test procedure is developed by modifying the weights used in the traditional repeated significance two-sample mean test. The new test has the type I error probability preserved at the target level and can provide a substantial gain in power with the increase of sample size. Generalization of the new procedure is discussed.     

Determination of nitrie in human blood by combination of a specific sample preparation with high-performance anion-exchange chromatography and electrochemical detection

All photometric or HPLC methods described to date have been unable to detect nitrite, a reliable marker of NO synthase activity, in human blood because of its rapid metabolism within the erythrocytes. We now elaborate on method to prevent nitrite degradation during sample preparation which in combination with high-performance anion-exchange chromatography and electrochemical detection allows a sensitive measurement of nitrite. A linear current response in the concentration range of 10–1000 nmol/l nitrite was observed yielding a correlation coefficient of 0.99. In addition, the combination of the electrochemical with a UV detector allowed us to simultaneously quantify nitrate one analytical run, which is the end product of NO/nitrite metabolism. Basal levels for nitrate and nitrite in human blood were determined with 25±4 μmol/l and 578±116 nmol/l (n=8), respectively and thus were in the same concentration range as expected from NO measurement in saline perfused isolated organs or cultured endothelial cells. Therefore, the presented method may be used to assess activity of endothelial constitutive NO synthase in humans under physiological and pathophysiological conditions.     

Effect of different sample pretreatment methods on the concentrations of excitatory amino acids in cerebrospinal fluid determined by high-performance liquid chromatography

Using high-performance liquid chromatography (HPLC), the effect of different sample pretreatment methods on the concentrations of excitatory amino acids (EAAs, glutamate and aspartate) measured in cerebrospinal fluid (CSF) was investigated. The results showed that the measured values of glutamate and aspartate were constant when the samples were stored at -80 degrees C and then methanol was used for CSF deproteinization before assay; the values of glutamate (Glu) increased when 0.3 M perchloric acid was used for CSF deproteinization with the CSF subsequently being stored at -20 degrees C; the values of Glu changed when the samples were stored at -20 degrees C over 8 weeks with methanol subsequently being used for CSF deproteinization before assay. This reference data suggested that the CSF sample would be better stored at -80 degrees C. If the sample is stored at -20 degrees C over 8 weeks, the Glu values change with the storage time. If strong acidic reagents are used for precipitation of protein in the CSF sample and then stored at -20 degrees C, Glu values are abnormally increased. From this study, an accurate and sensitive reversed-phase high-performance liquid chromatographic method has been developed for anti-excitotoxicity therapy and thorough study of EAAs in a clinical setting.     

Semi-automated liquid--liquid back-extraction in a 96-well format to decrease sample preparation time for the determination of dextromethorphan and dextrorphan in human plasma

A semi-automated, 96-well based liquid-liquid back-extraction (LLE) procedure was developed and used for sample preparation of dextromethorphan (DEX), an active ingredient in many over-the-counter cough formulations, and dextrorphan (DOR), an active metabolite of DEX, in human plasma. The plasma extracts were analyzed by liquid chromatography-tandem mass spectrometry (LC-MS-MS). The analytes were isolated from human plasma using an initial ether extraction, followed by a back extraction from the ether into a small volume of acidified water. The acidified water isolated from the back extraction was analyzed directly by LC-MS-MS, eliminating the need for a dry down step. A liquid handling system was utilized for all aspects of liquid transfers during the LLE procedure including the transfer of samples from individual tubes into a 96-well format, preparation of standards, addition of internal standard and the addition and transfer of the extraction solvents. The semi-automated, 96-well based LLE procedure reduced sample preparation time by a factor of four versus a comparable manually performed LLE procedure.     

Relative importance of within-habitat environment,land use and spatial autocorrelations for determining odonate assemblages in rural reservoir ponds in Japan

To clarify the major factors affecting odonate assemblages in rural reservoir ponds among within-habitat environments, land use around ponds and spatial autocorrelation, we surveyed odonate adults (Zygoptera and Anisoptera) in 70 study ponds in Ibaraki Prefecture, Japan, during three sampling periods in 2005. Redundancy analyses (RDA) for these three factor groups were executed to determine their strength in explaining the odonate species composition. Their relative contributions were also evaluated by the method of variation partitioning. A total of 41 odonate species were recorded in the study ponds, and 24 of them, excluding rare species, were used for our analysis. Summed effects including all three factor groups explained approximately 39% of the variation in odonate species composition. We found that spatial autocorrelation was the most important, though the within-habitat environment and land use had comparable effects. We conclude that spatial autocorrelation should be considered in this type of analysis, though we could not clearly explain what caused such a spatial structure. Pond area and debris that had accumulated at the bottom of ponds were selected as the within-habitat environment, and the forests and paddy fields around ponds were selected for land use after the procedure of forward stepwise selection. These results suggest that the recent decrease of forests around the ponds has had a negative effect on the odonate assemblages.     

Adaptive sample size calculations in group sequential trials

A method for group sequential trials that is based on the inverse normal method for combining the results of the separate stages is proposed. Without exaggerating the Type I error rate, this method enables data-driven sample size reassessments during the course of the study. It uses the stopping boundaries of the classical group sequential tests. Furthermore, exact test procedures may be derived for a wide range of applications. The procedure is compared with the classical designs in terms of power and expected sample size.     

Estimating the probability that the sample mean is within a desired fraction of the standard deviation of the true mean

The use of small sample sizes in human and primate evolutionary research is commonplace. Estimating how well small samples represent the underlying population, however, is not commonplace. Because the accuracy of determinations of taxonomy, phylogeny, and evolutionary process are dependant upon how well the study sample represents the population of interest, characterizing the uncertainty, or potential error, associated with analyses of small sample sizes is essential. We present a method for estimating the probability that the sample mean is within a desired fraction of the standard deviation of the true mean using small (n < 10) or very small (n ≤ 5) sample sizes. This method can be used by researchers to determine post hoc the probability that their sample is a meaningful approximation of the population parameter. We tested the method using a large craniometric data set commonly used by researchers in the field. Given our results, we suggest that sample estimates of the population mean can be reasonable and meaningful even when based on small, and perhaps even very small, sample sizes.