Search

[Display omitted] • Span 85, Span 80, Span 20, OP-4, and AOT greatly decreased the n -decane–water interfacial tension. • Span 85, Span 80, Span 20, OP-4, and AOT enhanced the spreading performance of n -decane droplets. • A system measuring the spreading performance of oil droplets containing surfactants on the air–water interface was built. • The law governing the spreading of surfactant-laden oil droplets on the surface of water has been systematically studied. • Surfactants in oil droplets migrated and were adsorbed at the oil–water interface, which was the mechanism of spreading. This study clarified the high-efficiency mechanism of enhancement of the spreading performance of oil by allowing a droplet of an organic solvent to spread over the surface of water with the addition of oil-soluble surfactants. Firstly, the effect of Span 85, Span 80, Span 20, octylphenol polyoxyethylene (4) ether (OP-4), and sodium dioctyl sulfosuccinate (AOT) on the interfacial parameters of an n -decane droplet was evaluated by measuring the surface tension and interfacial tension. Then the spreading coefficient was calculated to determine whether the surfactant-laden oil droplets could spread at the air–water interface. Moreover, the law governing the spreading of the surfactant-laden oil droplets over water was systematically studied. Furthermore, the characteristics of the dynamic behavior of the surfactants were investigated by molecular dynamics simulations to elucidate the spreading mechanism. The results of the experiments and simulations explained how the oil-soluble surfactants Span 85, Span 80, Span 20, OP-4, and AOT significantly improved the spreading performance of oil droplets over the water surface. [ABSTRACT FROM AUTHOR]

Previous studies demonstrate that part-of-speech (POS) tags are helpful for punctuation restoration tasks. However, extra computation cost will be needed during decoding, due to POS tags are provided by an external POS tagger. This paper proposes to transfer knowledge via adversarial training and orthogonality constraints to fill in the gap. Adversarial multi-task learning is introduced to learn task invariant knowledge from the extra POS tagging task for a punctuation prediction task. Furthermore, orthogonality constraints are used to make private and shared features dissimilar. Only the punctuation predicting task is used during decoding. So extra computation is not needed. Experiments are conducted on IWSLT2011 datasets. The results show that the punctuation predicting models trained with adversarial learning obtain performance gains over the baseline models on test sets. The results also demonstrate that the models trained with orthogonality constraints further obtain performance improvement. • Adversarial multi-task learning is used to learn knowledge from a POS tagging task for a punctuation prediction task. • Orthogonality constraints are used to make private and shared features dissimilar. • Extra computation is not needed, due to only the punctuation predicting task is used during decoding. • The punctuation predicting models trained with adversarial learning obtain performance gains over the baseline models. • The punctuation predicting models trained with orthogonality constraints further obtain performance improvement. [ABSTRACT FROM AUTHOR]

Many datasets have been designed to further the development of fake audio detection. However, fake utterances in previous datasets are mostly generated by altering timbre, prosody, linguistic content or channel noise of original audio. These datasets leave out a scenario, in which the acoustic scene of an original audio is manipulated with a forged one. It will pose a major threat to our society if some people misuse the manipulated audio with malicious purpose. Therefore, this motivates us to fill in the gap. This paper proposes such a dataset for scene fake audio detection named SceneFake, where a manipulated audio is generated by only tampering with the acoustic scene of an real utterance by using speech enhancement technologies. Some scene fake audio detection benchmark results on the SceneFake dataset are reported in this paper. In addition, an analysis of fake attacks with different speech enhancement technologies and signal-to-noise ratios are presented in this paper. The results indicate that scene fake utterances cannot be reliably detected by baseline models trained on the ASVspoof 2019 dataset. Although these models perform well on the SceneFake training set and seen testing set, their performance is poor on the unseen test set. The dataset 2 2 https://zenodo.org/record/7663324#.Y%5fXKMuPYuUk. and benchmark source codes 3 3 https://github.com/ADDchallenge/SceneFake. are publicly available. • This paper proposes a new problem: scene fake audio detection. • This is the first attempt to pose such an audio fake attack using speech enhancement. • This paper designs a dataset and provides benchmarks for scene fake audio detection. • The dataset provides speech enhancement technology information of fake utterances. • The dataset and benchmark source codes are publicly available. • The dataset will further foster research on fake audio detection. [ABSTRACT FROM AUTHOR]

• Handwriting style features are extracted by a Convolution Neural Network. • Calligraphy embedding is computed by attention and orthogonal mata-style matrix. • Calligraphy and character embedding are the inputs of Dual-condition Gated Recurrent Unit. • Handwriting Imitation results are generated by Gaussian Mixture Model and sampling. • Subjective and quantitative tests verify the imitation performance. Calligraphy imitation (CI) from a handful of target handwriting samples is such a challenging task that most of the existing writing style analysis or handwriting generation methods do not exhibit satisfactory performance. In this paper, we propose a novel multi-module framework to address the problem of CI. Firstly, we utilized a deep convolution neural network (CNN) to extract personalized calligraphical features. Then we built a calligraphy-clustering attention module and a mata-style matrix (msM) to compute an embedding of calligraphy. The structure of conditional gated recurrent unit (cGRU) is then improved to predict the probabilistic density of pen tip movement displacement by dual condition inputs. Finally, we generated personalized handwriting stroke sequences through iterative sampling with Gaussian mixture model (GMM). Experiments on public online handwriting databases verify that the proposed method could achieve satisfactory performance; the generated samples achieved high similarities with original handwriting examples. [ABSTRACT FROM AUTHOR]