scholarly journals An Effective Conversion of Visemes to Words for High-Performance Automatic Lipreading

Sensors ◽  
2021 ◽  
Vol 21 (23) ◽  
pp. 7890
Author(s):  
Souheil Fenghour ◽  
Daqing Chen ◽  
Kun Guo ◽  
Bo Li ◽  
Perry Xiao

As an alternative approach, viseme-based lipreading systems have demonstrated promising performance results in decoding videos of people uttering entire sentences. However, the overall performance of such systems has been significantly affected by the efficiency of the conversion of visemes to words during the lipreading process. As shown in the literature, the issue has become a bottleneck of such systems where the system’s performance can decrease dramatically from a high classification accuracy of visemes (e.g., over 90%) to a comparatively very low classification accuracy of words (e.g., only just over 60%). The underlying cause of this phenomenon is that roughly half of the words in the English language are homophemes, i.e., a set of visemes can map to multiple words, e.g., “time” and “some”. In this paper, aiming to tackle this issue, a deep learning network model with an Attention based Gated Recurrent Unit is proposed for efficient viseme-to-word conversion and compared against three other approaches. The proposed approach features strong robustness, high efficiency, and short execution time. The approach has been verified with analysis and practical experiments of predicting sentences from benchmark LRS2 and LRS3 datasets. The main contributions of the paper are as follows: (1) A model is developed, which is effective in converting visemes to words, discriminating between homopheme words, and is robust to incorrectly classified visemes; (2) the model proposed uses a few parameters and, therefore, little overhead and time are required to train and execute; and (3) an improved performance in predicting spoken sentences from the LRS2 dataset with an attained word accuracy rate of 79.6%—an improvement of 15.0% compared with the state-of-the-art approaches.

Materials ◽  
2020 ◽  
Vol 13 (9) ◽  
pp. 2156 ◽  
Author(s):  
Byeong Hoon Bae ◽  
Jeong Woo Lee ◽  
Jae Min Cha ◽  
Il-Won Kim ◽  
Hyun-Do Jung ◽  
...  

Powder bed fusion (PBF) additive manufacturing (AM) is currently used to produce high-efficiency, high-density, and high-performance products for a variety of applications. However, existing AM methods are applicable only to metal materials and not to high-melting-point ceramics. Here, we develop a composite material for PBF AM by adding Al2O3 to a glass material using laser melting. Al2O3 and a black pigment are added to a synthesized glass frit for improving the composite strength and increased laser-light absorption, respectively. Our sample analysis shows that the glass melts to form a composite when the mixture is laser-irradiated. To improve the sintering density, we heat-treat the sample at 750 °C to synthesize a high-density glass frit composite. As per our X-ray diffraction (XRD) analysis to confirm the reactivity of the glass frit and Al2O3, we find that no reactions occur between glass and crystalline Al2O3. Moreover, we obtain a high sample density of ≥95% of the theoretical density. We also evaluate the composite’s mechanical properties as a function of the Al2O3 content. Our approach facilitates the manufacturing of ceramic 3D structures using glass materials through PBF AM and affords the benefits of reduced process cost, improved performance, newer functionalities, and increased value addition.


Catalysts ◽  
2020 ◽  
Vol 10 (7) ◽  
pp. 802
Author(s):  
Chang Sun ◽  
Yingxin Mu ◽  
Yuxin Wang

Electrochemical ammonia synthesis, which is an alternative approach to the Haber–Bosch process, has attracted the attention of researchers because of its advantages including mild working conditions, environmental protection, and simple process. However, the biggest problem in this field is the lack of high-performance catalysts. Here, we report high-efficiency electroreduction of N2 to NH3 on γ-MnO2-supported Pd nanoparticles (Pd/γ-MnO2) under ambient conditions, which exhibits excellent catalytic activity with an NH3 yield rate of 19.72 μg·mg−1Pd h−1 and a Faradaic efficiency of 8.4% at −0.05 V vs. the reversible hydrogen electrode (RHE). X-ray diffraction (XRD) and transmission electron microscopy (TEM) characterization shows that Pd nanoparticles are homogeneously dispersed on the γ-MnO2. Pd/γ-MnO2 outperforms other catalysts including Pd/C and γ-MnO2 because of its synergistic catalytic effect between Pd and Mn.


Author(s):  
GÖKHAN YURTALAN ◽  
MURAT KOYUNCU ◽  
ÇİĞDEM TURHAN

Sentiment analysis attempts to resolve the senses or emotions that a writer or speaker intends to send across to the people about an object or event. It generally uses natural language processing and/or artificial intelligence techniques for processing electronic documents and mining the opinion specified in the content. In recent years, researchers have conducted many successful sentiment analysis studies for the English language which consider many words and word groups that set emotion polarities arising from the English grammar structure, and then use datasets to test their performance. However, there are only a limited number of studies for the Turkish language, and these studies have lower performance results compared to those studies for English. The reasons for this can be incorrect translation of datasets from English into Turkish and ignoring the special grammar structures in the latter. In this study, special Turkish words and linguistic constructs which affect the polarity of a sentence are determined with the aid of a Turkish linguist, and an appropriate lexicon-based polarity determination and calculation approach is introduced for this language. The proposed methodology is tested using different datasets collected from Twitter, and the test results show that the proposed system achieves better accuracy than the previously developed lexical-based sentiment analysis systems for Turkish. The authors conclude that especially analysis of word groups increases the overall performance of the system significantly.


2018 ◽  
Vol 3 (3) ◽  
pp. 582-597 ◽  
Author(s):  
Patrick Lemaire ◽  
Fleur Brun

The present study investigated how elementary-school children solve two-digit addition problems (e.g., 34+68). To achieve this end, we examined age-related differences in children’s strategy use and strategy performance. Results showed that (a) both third and fifth graders used a set of 9 strategies, (b) fifth-grade individuals used more strategies than third-grade individuals, (c) age-related differences in the size of strategy repertoire was partially explained by age-related differences in basic arithmetic fluency, (d) how often children used each available strategy changed with problem difficulty and children’s age, as younger children tended to focus more on one or two strategies and older children used a wider range of strategies, (e) increased arithmetic performance with age varied with problem difficulty both when overall performance was analyzed and when analyses of performance was restricted to children’s favorite strategy. The present findings have important implications for our understanding of how complex arithmetic performance changes with children’s age and change mechanisms underlying improved performance with age in complex arithmetic.


Author(s):  
Syed Simran.K, Md.Anwar, Dr.Sharan Reddy and Santosh.B.M

A high frequency DC-DC converter operating in the MHz range is proposed, which can achieve unbiased load current even while maintaining high performance over a wide range of load voltages. Due to these functions, the provided transformer is suitable for LED driver applications, which require different types of LEDs to operate with controlled current. The proposed transformer satisfies the uncontrolled load current using the LCL-T resonance community and achieves high efficiency using a predetermined switching frequency. The LCL-T resonance transformer also works effectively in controlling its output to the required rating using phase shift control. The overall performance of the LCL-T Echo transducer was evaluated and compared with the LC3L controlled echo transducer. Simulation work is done using the MATLAB / Simulink program.


Author(s):  
Mohamed Yousuf Hasan ◽  
Poornima V.P ◽  
Sujendran S ◽  
Karthikraja D

<p>This paper describes the design of high performance packet filtering firewall using embedded system. An FPGA (field programmable gate array) platform has been used for implementation and analysing the network firewall. It is capable of accepting real time changes. This network security application has an ability to perform powerful protection against unwanted data packets such as virus attack, spam in e-mails, hackers, worms, spyware unauthorized contents. However the firewalls don’t address the difficulty of unwanted data packets intrusion. The ultimate aim of this work is to create a systematic way of approach for unwanted packets discard in a network system. We use a specially trained algorithms such as Wu-manber algorithms (high performance, multi-pattern matching), bloom filter algorithm (space efficient data structure for testing an element in the set.Our design is mainly based on machine learning and artificial intelligence. This gives a high efficiency, improved performance and high ability of packet detection with less contribution of time in an effective way.</p>


2016 ◽  
Vol 11 (9) ◽  
pp. 764
Author(s):  
Lella Aicha Ayadi ◽  
Nihel Neji ◽  
Hassen Loukil ◽  
Mouhamed Ali Ben Ayed ◽  
Nouri Masmoudi

Pharmaceutics ◽  
2020 ◽  
Vol 12 (11) ◽  
pp. 1130
Author(s):  
Mariana Pires Figueiredo ◽  
Ana Borrego-Sánchez ◽  
Fátima García-Villén ◽  
Dalila Miele ◽  
Silvia Rossi ◽  
...  

This work presents the development of multifunctional therapeutic membranes based on a high-performance block copolymer scaffold formed by polyether (PE) and polyamide (PA) units (known as PEBA) and layered double hydroxide (LDH) biomaterials, with the aim to study their uses as wound dressings. Two LDH layer compositions were employed containing Mg2+ or Zn2+, Fe3+ and Al3+ cations, intercalated with chloride anions, abbreviated as Mg-Cl or Zn-Cl, or intercalated with naproxenate (NAP) anions, abbreviated as Mg-NAP or Zn-NAP. Membranes were structurally and physically characterized, and the in vitro drug release kinetics and cytotoxicity assessed. PEBA-loading NaNAP salt particles were also prepared for comparison. Intercalated NAP anions improved LDH–polymer interaction, resulting in membranes with greater mechanical performance compared to the polymer only or to the membranes containing the Cl-LDHs. Drug release (in saline solution) was sustained for at least 8 h for all samples and release kinetics could be modulated: a slower, an intermediate and a faster NAP release were observed from membranes containing Zn-NAP, NaNAP and Mg-NAP particles, respectively. In general, cell viability was higher in the presence of Mg-LDH and the membranes presented improved performance in comparison with the powdered samples. PEBA containing Mg-NAP sample stood out among all membranes in all the evaluated aspects, thus being considered a great candidate for application as multifunctional therapeutic dressings.


Processes ◽  
2021 ◽  
Vol 9 (6) ◽  
pp. 1034
Author(s):  
Ching-Chien Huang ◽  
Chin-Chieh Mo ◽  
Guan-Ming Chen ◽  
Hsiao-Hsuan Hsu ◽  
Guo-Jiun Shu

In this work, an experiment was carried out to investigate the preparation condition of anisotropic, Fe-deficient, M-type Sr ferrite with optimum magnetic and physical properties by changing experimental parameters, such as the La substitution amount and little additive modification during fine milling process. The compositions of the calcined ferrites were chosen according to the stoichiometry LaxSr1-xFe12-2xO19, where M-type single-phase calcined powder was synthesized with a composition of x = 0.30. The effect of CaCO3, SiO2, and Co3O4 inter-additives on the Sr ferrite was also discussed in order to obtain low-temperature sintered magnets. The magnetic properties of Br = 4608 Gauss, bHc = 3650 Oe, iHc = 3765 Oe, and (BH)max = 5.23 MGOe were obtained for Sr ferrite hard magnets with low cobalt content at 1.7 wt%, which will eventually be used as high-end permanent magnets for the high-efficiency motor application in automobiles with Br > 4600 ± 50 G and iHc > 3600 ± 50 Oe.


2021 ◽  
Vol 12 (11) ◽  
pp. 1692-1699
Author(s):  
Ji Hye Lee ◽  
Jinhyo Hwang ◽  
Chai Won Kim ◽  
Amit Kumar Harit ◽  
Han Young Woo ◽  
...  

New polystyrene-based polymers with high π-extended hole transport pendants were synthesized to obtain a low turn-on voltage and high efficiency in solution-processed green TADF-OLEDs.


Sign in / Sign up

Export Citation Format

Share Document