A5091502081- Microstructure and Mechanical Properties of Steels
- Metal Alloys Wear and Properties
- Speech Recognition and Synthesis
- Music and Audio Processing
- Speech and Audio Processing
- Risk and Safety Analysis
- Occupational Health and Safety Research
- Metal and Thin Film Mechanics
- Hydrogen embrittlement and corrosion behaviors in metals
- Natural Language Processing Techniques
- Advanced materials and composites
- Magnetic Properties and Applications
- User Authentication and Security Systems
- Topic Modeling
- Reliability and Maintenance Optimization
- Railway Engineering and Dynamics
- Microstructure and mechanical properties
- Electrospun Nanofibers in Biomedical Applications
- Multi-Criteria Decision Making
- Rice Cultivation and Yield Improvement
- Agricultural Science and Fertilization
- Traffic and Road Safety
- Corrosion Behavior and Inhibition
- Network Security and Intrusion Detection
- Customer churn and segmentation
Amity University
2024
Helmholtz-Zentrum Hereon
2022-2024
Indian Institute of Technology Delhi
2024
Indian Institute of Technology Kharagpur
2018-2022
Indian Institute of Technology Bombay
2018-2022
Narendra Dev University of Agriculture and Technology
2022
Meta (United States)
2021
National Institute of Technology Raipur
2021
Meta (Israel)
2019-2020
Banaras Hindu University
2014-2019
This paper presents XLS-R, a large-scale model for cross-lingual speech representation learning based on wav2vec 2.0.We train models with up to 2B parameters nearly half million hours of publicly available audio in 128 languages, an order magnitude more public data than the largest known prior work.Our evaluation covers wide range tasks, domains, regimes and both high low-resource.On CoVoST-2 translation benchmark, we improve previous state art by average 7.4 BLEU over 21 directions into...
Language identification greatly impacts the success of downstream tasks such as automatic speech recognition. Recently, self-supervised representations learned by wav2vec 2.0 have been shown to be very effective for a range tasks. We extend previous work on language experimenting with pre-trained models which were real-world unconstrained in multiple languages and not just English. show that many perform better enable systems require little labeled data well. Results 26 setup only 10 minutes...
Although speaker verification has conventionally been an audio-only task, some practical applications provide both audio and visual streams of input. In these cases, the stream provides complementary information can often be leveraged in conjunction with acoustics speech to improve performance. this study, we explore audio-visual approaches verification, starting standard fusion techniques learn joint (AV) embeddings, then propose a novel approach handle cross-modal at test time....
β-Amylase finds application in food and pharmaceutical industries. Functionalized graphene sheets were customised as a matrix for covalent immobilization of Fenugreek β-amylase using glutaraldehyde cross-linker. The factors affecting the process optimized Response Surface Methodology based Box-Behnken design experiment which resulted 84% efficiency. Scanning Transmission Electron Microscopy (SEM, TEM) Fourier Tansform Infrared (FTIR) spectroscopy employed purpose characterization attachment...
Towards developing high-performing ASR for low-resource languages, approaches to address the lack of resources are make use data from multiple and augment training by creating acoustic variations. In this work we present a single grapheme-based model learned on 7 geographically proximal using standard hybrid BLSTM-HMM models with lattice-free MMI objective. We build grapheme set via taking union over each language-specific set, find such multilingual graphemic can perform...
Many semi- and weakly-supervised approaches have been investigated for overcoming the labeling cost of building high quality speech recognition systems. On challenging task transcribing social media videos in low-resource conditions, we conduct a large scale systematic comparison between two self-labeling methods on one hand, pretraining using contextual metadata other. We investigate distillation at frame level sequence hybrid, encoder-only CTC-based, encoder-decoder systems Dutch Romanian...
Abstract Carbide free nano-bainitic (NSB) steels have a dual phase structure comprising of retained austenite (RA) and bainitic ferrite (BF). The content morphology RA BF in NSB can be easily modulated by changing the austempering temperature. This results significant changes strength is expected to influence corrosion resistance as well. behavior three steel blocks austempered at 250, 300 350 °C has been studied aqueous chloride (3.5 wt% NaCl) environment. Electrochemical impedance...
Strength of carbide-free bainitic steels primarily depends on the thickness ferrite and its volume fraction in microstructure. The morphology phases formed transformation are governed by steel's composition temperature. Higher carbon concentration is expected to refine microstructure even when austempering temperature kept be around 300 °C due enhanced strength austenite. However, there a concomitant loss bainite formed. Therefore, predicting optimal for getting highest hardness...