- Speech and Audio Processing
- Speech Recognition and Synthesis
- Music and Audio Processing
- Food composition and properties
- Polysaccharides Composition and Applications
- Phytochemicals and Antioxidant Activities
- Phytase and its Applications
- Natural Language Processing Techniques
- Blind Source Separation Techniques
- Advanced Adaptive Filtering Techniques
- Nanocomposite Films for Food Packaging
- Seed and Plant Biochemistry
- Microbial Metabolites in Food Biotechnology
- Proteins in Food Systems
- Nanoparticles: synthesis and applications
- Microbial Inactivation Methods
- Retinal and Macular Surgery
- Dyeing and Modifying Textile Fibers
- Spectroscopy and Chemometric Analyses
- Freezing and Crystallization Processes
- Food Drying and Modeling
- biodegradable polymer synthesis and properties
- Intraocular Surgery and Lenses
- Geophysical Methods and Applications
- Music Technology and Sound Studies
Sardar Patel University
2023-2024
The Charutar Vidya Mandal (CVM) University
2023-2024
Indian Institute of Technology Hyderabad
2023-2024
Intel (India)
2024
PES University
2023
Microsoft Research (United Kingdom)
2018-2022
Creative Commons
2022
Infant
2022
National Institute of Technology Raipur
2021
Microsoft (United States)
2009-2021
Almost all current automatic speech recognition (ASR) systems conventionally append delta and double-delta cepstral features to static features. In this work we describe a modified feature-extraction procedure in which the time-difference operation is performed spectral domain, rather than domain as generally presently done. We argue that approach based on "delta-spectral" needed because even though delta-cepstral capture dynamic information greatly improve ASR accuracy, they are not robust...
The present study focused on impact of microwave heat treatment the morphological, functional, thermal, and antioxidant properties quinoa. Microwave roasting significantly (p < .05) varied WAC (1.04 to 1.41 g/g), OAC (1.09 0.96 FC (6.66% 4.62%), EC (57.06% 43.18%), pasting Roasted quinoa samples are rich in TPC (78.54 105.9 mg GAE/100 g) AOA (68.95% 81.84%). In contrast, higher reduction gelatinization temperature destruction crystalline structure (amylopectin content: 85.23% 80.78%) starch...
In this paper, we introduce profile view (PV) lip reading, a scheme for speaker-dependent isolated word speech recognition. We provide historic motivation PV from the importance of images in facial animation and present feature extraction schemes as well traditional frontal (FV) approach. compare reading results FV, which demonstrate significant improvement over FV. show recognition with integration audio visual features. also found it advantageous to process features longer duration than...
Recently Long Short-Term Memory (LSTM) Recurrent Neural Networks (RNN) acoustic models have demonstrated superior performance over deep neural networks (DNN) in speech recognition and many other tasks. Although a lot of work been reported on DNN model adaptation, very little has done LSTM adaptation. In this paper we present our extensive studies speaker adaptation LSTM-RNN for recognition. We investigated different methods combined with KL-divergence based regularization, where which...
Purpose The amount of food wasted every year is 1.3 billion metric tonne (MT), out which 0.5 MT contributed by the fruits processing industries. waste includes by-products such as peels, pomace and seeds a good source bioactive compounds like phenolic compounds, flavonoids, pectin lipids dietary fibres. Hence, purpose present study to review novel extraction techniques used for bio active from selection suitable method. Design/methodology/approach Novel ultrasound-assisted extraction,...
Abstract The effects of different concentrations soy protein isolate (SPI) on the functional, rheological, pasting, and structural properties potato starch (PS) are investigated. water oil absorption capacity (OAC) PS‐SPI mix increased significantly with addition SPI. Gel strength, peak, final, breakdown, setback viscosity PS decreases, whereas pasting temperature increases an increase in SPI fraction. Both Herschel–Bulkley power law model found to equally fit steady state rheology. blend...
Purpose This study aims to fabricate IN718 parts through laser-powder bed fusion (L-PBF) at varied scanning speeds. In the as built condition, studies on location-specific microstructure evolution followed by characterization of mechanical properties such tensile (ultimate strength and % elongation) microhardness are carried out. The obtained correlated with part porosity/relative density. specimens then post-heat treated (homogenization treatment double-stage aging). compared that well...
We present an algorithm for dereverberation of speech signals automatic recognition (ASR) applications. Often ASR systems are presented with that has been recorded in environments include noise and reverberation. The performance degrades increasing levels While many algorithms have proposed robust noisy environments, reverberation is still a challenging problem. In this paper, we ' approach models as convolution operation the spectral domain. Using least-squares error criterion decompose...
In this paper, we present a new two-microphone approach that improves speech recognition accuracy when is masked by other speech. The algorithm on previous systems have been successful in separating signals based differences arrival time of signal components from two microphones. differs these efforts the selection takes place frequency domain. We observe additional smoothing phase estimates over and needed to support adequate performance. demonstrate described paper provides better than...
Confidence classifiers for automatic speech recognition (ASR) provide a quantitative representation the reliability of ASR decoding. In this paper, we improve confidence measure performance an utterance using two distinct approaches: (1) to define and incorporate additional predictors in classifier including those based on word identity aggregated words, (2) train built deep learning architectures neural network (DNN) kernel convex (K-DCN). Our experiments show that adding new our...
Water hydration characteristics of moong at temperatures 25, 35, 45 and 55 °C were determined three major dimensions (length, width thickness) moisture content time interval varying from 0 to 500 min. Peleg's model Fick's second law diffusion used simulate water absorption effective diffusivity with temperature. equation was adequately capable predicting the experimental condition rate constants k1 k2 decreased increased soaking temperature 0.5132 0.1081 0.0165 0.151, respectively. Effective...
In the presence of a wide variety dialects, training dialect-specific models for each dialect is demanding task. Previous studies have explored single model that robust across multiple dialects. These used either multi-condition training, multi-task learning, end-to-end modeling, or ensemble modeling. this study, we further explore using multi-dialect speech recognition First, build an (or experts). Then linearly combine outputs experts attention weights generated by long short-term memory...
Microorganisms under natural habitats live in communities and some provide benefits to plants. The concept of development microbial consortia for bio-control crop sustainability relies on this fact. Microbes when introduced soil as consortium interact with a host plant, partially mimic the conditions. To improve stability released microbes different agricultural fields, use is advocated. Further, together can also offer multiple mechanisms mycoparasitism, competition, antibiosis, induced...