A stacked self-attention network for two-dimensional direction-of-arrival estimation in hands-free speech communication.

When making voice interactions with hands-free speech communication devices, direction-of-arrival estimation is an essential step. To address the detrimental influence of unavoidable background noise and interference speech on direction-of-arrival estimation, this paper introduces a stacked self-attention network system, a supervised deep learning method that enables utterance level estimation without requirement for any pre-processing such as voice activity detection. Specifically, alternately stacked time- and frequency-dependent self-attention blocks are designed to process information in terms of time and frequency, respectively. The former blocks focus on the importance of each time frame of the received audio mixture and perform temporal selection to reduce the influence of non-speech and interference frames, while the latter blocks are utilized to derive inner-correlation among different frequencies. Additionally, the non-causal convolution and self-attention networks are replaced by causal ones, enabling real-time direction-of-arrival estimation with a latency of only 6.25 ms. Experiments with simulated and measured room impulse responses, as well as real recordings, verify the advantages of the proposed method over the state-of-the-art baselines.

Long-term scalogram integrated with an iterative data augmentation scheme for acoustic scene classification.

In acoustic scene classification (ASC), acoustic features play a crucial role in the extraction of scene information, which can be stored over different time scales. Moreover, the limited size of the dataset may lead to a biased model with a poor performance for recordings from unseen cities and confusing scene classes. This paper proposes a long-term wavelet feature that captures discriminative long-term scene information. The extracted scalogram requires a lower storage capacity and can be classified faster and more accurately compared with classic Mel filter bank coefficients (FBank). Furthermore, a data augmentation scheme is adopted to improve the generalization of the ASC systems, which extends the database iteratively with auxiliary classifier generative adversarial neural networks (ACGANs) and a deep learning-based sample filter. Experiments were conducted on datasets from the Detection and Classification of Acoustic Scenes and Events (DCASE) challenges. The DCASE17 and DCASE19 datasets marked a performance boost of the proposed techniques compared with the FBank classifier. Moreover, the ACGAN-based data augmentation scheme achieved an absolute accuracy improvement of 6.10% on recordings from unseen cities, far exceeding classic augmentation methods.

A dual-stream deep attractor network with multi-domain learning for speech dereverberation and separation.

Deep attractor networks (DANs) perform speech separation with discriminative embeddings and speaker attractors. Compared with methods based on the permutation invariant training (PIT), DANs define a deep embedding space and deliver a more elaborate representation on each time-frequency (T-F) bin. However, it has been observed that the DANs achieve limited improvement on the signal quality if directly deployed in a reverberant environment. Following the success of time-domain separation networks on the clean mixture speech, we propose a dual-stream DAN with multi-domain learning to efficiently perform both dereverberation and separation tasks under the condition of variable numbers of speakers. The speaker encoding stream (SES) of the dual-stream DAN is trained to model the speaker information in the embedding space defined with the Fourier transform kernels. The speech decoding stream (SDS) accepts speaker attractors from the SES and learns to estimate the early component of the sound in the time domain. Meanwhile, additional clustering losses are used to bridge the gap between the oracle and the estimated attractors. Experiments were conducted on the Spatialized Multi-Speaker Wall Street Journal (SMS-WSJ) dataset. After comparing with the anechoic and reverberant signals, the early component was chosen as the learning targets. The experimental results demonstrated that the dual-stream DAN achieved scale-invariant source-to-distortion ratio (SI-SDR) improvement of 9.8∕7.5 dB on the reverberant 2-/3-speaker evaluation set, exceeding the baseline DAN and convolutional time-domain audio separation network (Conv-TasNet) by 2.0∕0.7 dB and 1.0∕0.5 dB, respectively.

Fast speciation of mercury in seawater by short-column high-performance liquid chromatography hyphenated to inductively coupled plasma spectrometry after on-line cation exchange column preconcentration.

A simple and fast method for trace speciation analysis of mercury (Hg(2+)), methylmercury (MeHg(+)) and ethylmercury (EtHg(+)) in seawater has been developed by short-column high-performance liquid chromatography hyphenated to inductively coupled plasma spectrometry (HPLC-ICP-MS) after on-line cation-exchange column (CEC) preconcentration. The analytes were firstly adsorbed on the CEC without any extraneous reagent, and then were eluted rapidly (within seconds) and completely with a very low concentration of l-cysteine solution, which provides the conveniency for the on-line coupling of the preconcentration method and detection technique. To our best knowledge, it is for the first time to employ the CEC preconcentration technique to trap all of the three mercury species simultaneously at their positive charged status for the purpose of speciation analysis. Under the optimized conditions, a very high preconcentration factor up to 1250 has been obtained with 30mL sample solution, which leads to the very low detection limits of 0.042ngL(-1) for Hg(2+), 0.016ngL(-1) for MeHg(+) and 0.008ngL(-1) for EtHg(+) (as Hg), respectively. With the established method, three seawater samples were also analyzed, and all the three mercury species have been found in each sample, albeit at a very low concentration.

Dispersive solid-phase extraction combined with dispersive liquid-liquid microextraction for the determination of polybrominated diphenyl ethers in plastic bottled beverage by GC-MS.

A simple, inexpensive and reliable analytical method was developed for the determination of polybrominated diphenyl ethers (PBDEs) in polyethylene terephthalate (PET) bottled beverage using GC-MS. The sample pretreatment using dispersive solid-phase extraction (DSPE) for removing matrix and dispersive liquid-liquid microextraction (DLLME) for enriching analytes was performed. For the DSPE, different sorbents such as primary amine, secondary amine, C(18) and graphitized carbon black were tested for different sample matrices. By means of DSPE, 60-89% of the sample matrices could be removed. Acetonitrile solution obtained by DSPE cleanup was directly used as the dispersant for the subsequent DLLME, which made the combination of the DSPE with the DLLME much more straightforward. Under the optimal conditions, the enrichment factors (EFs) of PBDEs ranged from 199 to 292. Using matrix-matched calibration, correlation coefficients above 0.994 were found and LODs ranged from 0.0023 to 0.15 µg/L. The recoveries were between 80 and 117% for beverages spiked at three different concentrations (1.0, 5.0 and 10 µg/L) with RSDs ranging from 3.7 to 14.7% (n=5). The results indicated that the combination of DSPE with DLLME was a powerful sample preparation tool for analysis of ultratrace analytes in complicated matrices.

Influence of plant activity and phosphates on thorium bioavailability in soils from Baotou area, Inner Mongolia.

Harm of thorium to living organisms is governed by its bioavailability. Thorium bioavailability in the soil-plant system of Baotou rare earth industrial area was studied using pot experiments of wheat and single extraction methods. The effects of wheat growth stage and phosphate on thorium bioavailability were also investigated. Based on extractabilities of various extraction methods (CaCl(2), NH(4)NO(3), EDTA, HOAc) and correlation analysis of thorium uptake by wheat plant and extractable thorium, a mixture of 0.02M EDTA+0.5M NH(4)OAc (pH 4.6) was found suitable for evaluation of thorium bioavailability in Baotou soil, which could be predicted quantitatively by multiple regression models. Because of differences of wheat root activities, thorium bioavailability in rhizosphere soil was higher than in bulk soil at tillering stage, but the reverse occurred at jointing stage. Phosphate addition induced the mineralization of soluble thorium by forming stable thorium phosphate compounds, and reduced thorium bioavailability in soil.

[Adsorption behavior of exogenous thorium on soil contaminated by rare earth industries].

The adsorption behavior of exogenous thorium on soil was studied to evaluate the contaminated risk on soil. The adsorption capacity, equilibrium time, distribution coefficient and desorption ability were investigated by the experiments of static adsorption. The strong adsorption ability of exogenous thorium on soil samples was observed by high adsorption ratio (> 92%) and low desorption ratio (< 5%) in equilibrium, and the biggest distribution coefficient was over 10(4). The adsorption capacity and equilibrium time were related to soil properties. According to the results of adsorption, Freundlich equation (r > or = 0.9167) and Elovich equation (R2 > or = 0.8980) were primely fit for describing the thermodynamics and kinetics of the adsorption of exogenous thorium on soil samples, respectively, which indicated that the adsorption was belonged to the nonlinear adsorption, and was affected by the diffusion of thorium on soil surface and in mineral interbed. Sequential extraction procedure was employed to evaluate the bound fractions of exogenous thorium adsorbed on soil samples. Based on the extracted results of thorium fractions, exogenous thorium was presented in the labile nonresidual fractions (over 58%) at the low initial concentration (10(-7) - 10(-6) mol x L(-1)), and nonresidual fractions enhanced with the increase of the initial amount, meanwhile more exogenous throium was transferred to the stable residual fractions.

Evaluation of a sequential extraction for the speciation of thorium in soils from Baotou area, Inner Mongolia.

Sequential extraction procedures were widely applied for speciation of radioactive elements. In this study, the sequential extraction procedure developed by Martínez-Aguirre was employed for quantification of different chemical forms of thorium in the soil. The total amount of thorium in contaminated soil was much higher by four-fold than the local background value. The soil properties affect the amount of thorium and distribution of fractions in contaminated soil. Results showed that the proportion of thorium in soils from Baotou was found as the residual fraction (F5+F6)>absorbed fraction (F3), coprecipitated fraction (F4)>carbonates fraction (F2) and exchangeable fraction (F1) that could be available to plants. The recovery, calculated by ratio of the sum of the six fractions to the pseudo-total content of thorium, was in the range from 96% to 110%. A comparison was carried out between the sequential extraction and the single extraction to evaluate the selectivity of the extractants. It was found that the amount of thorium of absorbed fraction (F3) was higher in the single extraction than that estimated in the sequential extraction, possibly duo to transform of the labile form. While for non-residual fraction analysis, the single extraction scheme was a desirable alternative to the sequential extraction procedure. According to correlativity analysis of various fractions, it might be predicted that how the non-residual fractions of thorium were directionally transformed into interrelated fractions under the changes of conditions.

Preliminary results on the determination of ultratrace amounts of cadmium in tea samples using a flow injection on-line solid phase extraction separation and preconcentration technique to couple with a sequential injection hydride generation atomic fluorescence spectrometry.

In this work, a method was developed for determination of ultra-trace levels of Cd in tea samples by atomic fluorescence spectrometry (AFS). A flow injection solid phase extraction (FI-SPE) separation and preconcentration technique, to on-line couple with a sequential injection hydride generation (SI-HG) technique is employed in this study. Cd was preconcentrated on the SPE column, which was made from a neutral extractant named Cyanex 923, while other matrix ions or interfering ions were completely or mostly separated off. Conditions for the SPE separation and preconcentration, as well as conditions for the HG technique, were studied. Due to the separation of interfering elements, Cd hydride generation efficiency could be greatly enhanced with the sole presence of Co(2+) with a concentration of 200mugL(-1), which is much lower than those in other works previously reported. Interferences on both the Cd separation and preconcentration, and Cd hydride generation (HG) were investigated; it showed that both the separation and preconcentration system, and the HG system had a strong anti-interference ability. The SPE column could be repeatedly used at least 400 times, a R.S.D. of 0.97% was obtained for 6 measurements of Cd with 0.2mugL(-1) and a correlation coefficiency of 1.0000 was obtained for the measurement of a series of solutions with Cd concentrations from 0.1 to 2mugL(-1). The method has a low detection limit of 10.8ngL(-1) for a 25mL solution and was successfully validated by using two tea standard reference materials (GBW08513 and GBW07605).

Cyanex 923 as the extractant in a rare earth element impurity analysis of high-purity cerium oxide.

In this work, the feasibility of employing Cyanex 923 as an extractant into the non-cerium REE (rare earth elements) impurity analysis of high-purity cerium oxide was investigated. Through investigations on the choice of the extraction medium, the optimium extraction acidity, matrix Ce4+ effect on the non-cerium REE ion extraction, the optimium extractant concentration and suitable extracting time, and oscillation strengh, it was found that when the phase ratio was at 1:1 and the acicidity was about 2% H2SO4, by gently shaking by hand for about 2 min, 10 mL of 30% Cyanex 923 could not extract even for a 20 ng amount of non-cerium REE3+ ions. However, the extraction efficiency for Ce4+ of 100 mg total amount under the same conditions was about 96%, indicating that a 25-fold preconcentration factor could be achieved. Thus, it was concluded that Cyanex 923 could be used in a REE impurity analysis of 99.9999% or so pure cerium oxide for primary sepapation to elimilate matrix-induced interferences encountered in an ICP-MS (inductively coupled plasma mass spectroscopy) determination.

[Major, minor and trace-elemental contents analysis in northeast soybeans by ICP-AES].

Major, minor and trace elemental contents in northeast China soybeans were determined by using inductively coupled plasma atomic emission spectrometry (ICP-AES). Three different sample digestion methods including two wet digestions, HNO3-HClO4 and HNO3-H2SO4 and a dry ash method were compared. Owing to the high oil content in soybeans, long time is needed and access acid should be added with mixed acid digestion methods, which may result in higher sample blank. Therefore, the dry ask method would be more proper for the pre-treatment of soybean samples. Potassium and phosphorus are major elements in soybeans, so the effect of potassium and phosphorus on the other elements was investigated. Results showed that the potassium and phosphorus did not affect the determination of other trace elements. There are not significant differences in trace elemental contents for the eleven northeast China soybeans.