Communication has been an important aspect of human life, civilization, and globalization for thousands of years. Biometric analysis, education, security, healthcare, and smart cities are only a few examples of speech recognition applications. Most studies have mainly concentrated on English, Spanish, Japanese, or Chinese, disregarding other low-resource languages, such as Uzbek, leaving their analysis open. In this paper, we propose an End-To-End Deep Neural Network-Hidden Markov Model speech recognition model and a hybrid Connectionist Temporal Classification (CTC)-attention network for the Uzbek language and its dialects. The proposed approach reduces training time and improves speech recognition accuracy by effectively using CTC objective function in attention model training. We evaluated the linguistic and lay-native speaker performances on the Uzbek language dataset, which was collected as a part of this study. Experimental results show that the proposed model achieved a word error rate of 14.3% using 207 h of recordings as an Uzbek language training dataset.

/pdf/automatic-speech-recognition-method-based-on-deep-learning-2l9hjndw.pdf

Automatic Speech Recognition Method Based on Deep Learning Approaches for Uzbek Language

GAN review: Models and medical image fusion applications

The presented paper introduces principal component analysis application for dimensionality reduction of variables describing speech signal and applicability of obtained results for the disturbed and fluent speech recognition process. A set of fluent speech signals and three speech disturbances—blocks before words starting with plosives, syllable repetitions, and sound-initial prolongations—was transformed using principal component analysis. The result was a model containing four principal components describing analysed utterances. Distances between standardised original variables and elements of the observation matrix in a new system of coordinates were calculated and then applied in the recognition process. As a classifying algorithm, the multilayer perceptron network was used. Achieved results were compared with outcomes from previous experiments where speech samples were parameterised with the Kohonen network application. The classifying network achieved overall accuracy at 76% (from 50% to 91%, depending on the dysfluency type).

/pdf/artificial-neural-networks-combined-with-the-principal-1d6ewci0.pdf

Artificial Neural Networks Combined with the Principal Component Analysis for Non-Fluent Speech Recognition

Charm bracelets and their application to the construction of periodic Golay pairs

/pdf/the-art-of-computer-programming-volume-4-generating-all-14vpn9jz7c.pdf

The Art of Computer Programming, Volume 4: Generating All Trees, History of Combinatorial Generation.

Generative adversarial networks for speech processing: A review

http://www.cis.uoguelph.ca/~sawada/papers/fix-brace.pdf

Generating bracelets with fixed content

Federated recommenders: methods, challenges and future

Pseudo bicliques relax the rigid connectivity requirement of bicliques to effectively deal with missing data. In this paper, we propose an algorithm based on reverse search to generate all pseudo bicliques in a given graph. We introduce various enhancements to our algorithm based on the structure of pseudo bicliques and underlying bipartite graph. We perform composite analysis using theoretical bounds and computational experiments, to show that these improvements significantly reduce the running time of our algorithm. Our algorithm is optimal in the sense that it takes average linear time to generate each pseudo biclique.

On fast enumeration of pseudo bicliques

Pseudo-bicliques model various problems encountered in bio-informatics, data mining and networks. They relax the rigid connectivity requirement of bicliques to cater missing and noisy data. In this paper, we consider the weighted density based model of pseudo-biclique. This model defines pseudo-biclique as a bipartite subgraph such that the ratio of the number of its edges to the number of edges in biclique of same size is no less than a given threshold value. The weighted model of pseudo-bicliques better fits the real-world situations and give much more flexibility to researchers. We propose an algorithm based on reverse search to generate all weighted pseudo-bicliques in a given graph. This is computationally non-trivial task as simple straightforward branch-and-bound and back-tracking schemes involve an NP complete problem. Furthermore, we proposed the use of multi-level pseudo-bicliques to discover knowledge at multiple levels and extend our algorithm to enumerate all multi-level pseudo-bicliques. We introduce various enhancements to our algorithm based on the structure of pseudo-bicliques and underlying bipartite graph. We evaluated the performance of our algorithms on random graphs and real-world problems. The results are quite promising and show that average linear time is incurred to generate each pseudo-biclique.

Saira Karim

Papers

Generative adversarial networks for speech processing: A review

Generating bracelets with fixed content

Federated recommenders: methods, challenges and future

On fast enumeration of pseudo bicliques

Mining Dense Structures by Enumerating Weighted and Multi-level Pseudo-Bicliques