Search

Wav2vec 2.0 is a recently proposed self-supervised framework for speech representation learning. It follows a two-stage training process of pre-training and fine-tuning, and performs well in speech recognition tasks especially ultra-low resource cases. In this work, we attempt to extend self-supervised framework to speaker verification and language identification. First, we use some preliminary experiments to indicate that wav2vec 2.0 can capture the information about the speaker and language. Then we demonstrate the effectiveness of wav2vec 2.0 on the two tasks respectively. For speaker verification, we obtain a new state-of-the-art result, Equal Error Rate (EER) of 3.61% on the VoxCeleb1 dataset. For language identification, we obtain an EER of 12.02% on 1 second condition and an EER of 3.47% on full-length condition of the AP17-OLR dataset. Finally, we utilize one model to achieve the unified modeling by the multi-task learning for the two tasks., Self-supervised, speaker verification, language identification, multi-task learning, wav2vec 2.0

Recently, end-to-end (E2E) models become a competitive alternative to the conventional hybrid automatic speech recognition (ASR) systems. However, they still suffer from speaker mismatch in training and testing condition. In this paper, we use Speech-Transformer (ST) as the study platform to investigate speaker aware training of E2E models. We propose a model called Speaker-Aware Speech-Transformer (SAST), which is a standard ST equipped with a speaker attention module (SAM). The SAM has a static speaker knowledge block (SKB) that is made of i-vectors. At each time step, the encoder output attends to the i-vectors in the block, and generates a weighted combined speaker embedding vector, which helps the model to normalize the speaker variations. The SAST model trained in this way becomes independent of specific training speakers and thus generalizes better to unseen testing speakers. We investigate different factors of SAM. Experimental results on the AISHELL-1 task show that SAST achieves a relative 6.5% CER reduction (CERR) over the speaker-independent (SI) baseline. Moreover, we demonstrate that SAST still works quite well even if the i-vectors in SKB all come from a different data source other than the acoustic training set.

In order to find a model solution to simulate actual extracellular polymeric substances (EPS) solution in terms of filterability behavior, a series of experiments were conducted in a dead-end unstirred cell with 0.1 μm polyvinylidene fluoride membranes using single/binary mixtures consisting of sodium alginate (SA), bovine serum albumin (BSA) and humic acid (HA). It can obtain that the presence of BSA and SA is a necessary condition for model the filtration behavior of the actual EPS solution. Meanwhile, compared the actual EPS solution with binary mixtures model EPS solution can find that BSA/SA=140/58.3 and SA/HA=107/28.5 can exhibit the similar filtration behavior as actual EPS, while BSA/HA are far from that of actual EPS solution. The suitable type and combination of simulated solution were determined to improve the consistency with EPS solution in actual MBR system.

The purpose of this work was to investigate feasibility of a promising topical drug delivery system (TDDS) for sinomenine hydrochloride (SMH), extracted from the Chinese medicinal plant sinomenine acutum and currently used for the treatment of rheumatoid arthritis. It was found that SMH was a weak base (p K a , 7.98 ± 0.04) with pH‐dependent solubility and partition coefficient. The result of in vitro permeation studies demonstrated that the permeation enhancer azone was the most effective. In contrast, spray had higher accumulative permeated amounts of SMH than patch, but permeated duration of spray was shorter than that of patch. The efficacy on Freund's complete adjuvant‐induced arthritis suggested that there was near arthritis index for SMH spray with medium dose (i.e., 15 mg/rat) and oral solution at a dose of 12 mg/rat, indicating that topical SMH delivery system could achieve the similar anti‐inflammatory efficacy with oral administration. Pharmacokinetic parameters including C max and AUC for both topical preparations were lower than those for oral preparation, which hinted that systemic side effect could be ignored. Therefore, the spray and patch were promising formulations for successful topical delivery of SMH through the skin instead of oral administration with side effects. © 2009 Wiley‐Liss, Inc. and the American Pharmacists Association J Pharm Sci 99: 1790–1799, 2010

Disinfection for drinking water ought to be safe both in microbiology and in chemistry. This paper addresses by-products with carcinogenicity-mutagenicity-teratogenicity produced during chlorination disinfection. Types and formation of chlorination disinfection by-products (CDBPs) in drinking water are briefly analyzed. Toxicity and health effect of the CDBPs is reviewed. Measures to control CDBPs are proposed. And future direction in safe disinfection of drinking water is outlined thereby. Results showed that types and concentration of precursors and chlorine dosage are decisive factors in CDBPs formation; Water quality characteristics such as pH, temperature, bromide ions and ammonia, can also have an effect on the formation of CDBPs; Trihalomethanes (THMs), haloacetic acids(HAAs) and halogenated acetonitrile have carcinogenic potential or carcinogenicity. Protection of water resources is a fundamental measure to control CDBPs. Enhanced coagulation, activated carbon adsorption and membrane treatment can be used as the main technologies to remove CDBPs precursors. Optimizing disinfection process, using an alternative disinfectant or sequential disinfection mode are also important ways to control the CDBPs.